56: Belt and Suspenders


00:00:02   this is hypercritical weekly talkshow [TS]

00:00:04   ruminating on exactly what is wrong in [TS]

00:00:07   the world of Apple and related [TS]

00:00:08   technologies and businesses [TS]

00:00:10   nothing is so perfect that cannot be [TS]

00:00:13   destroyed by my co-host John siracusa hi [TS]

00:00:16   John hi Dan I'm Dan Benjamin this is [TS]

00:00:20   today's Friday February 24 2012 is [TS]

00:00:25   episode number 56 we've got two great [TS]

00:00:28   sponsors today apps feiyr.com and [TS]

00:00:31   MailChimp comm tell me more about them [TS]

00:00:34   as the program goes on we also want to [TS]

00:00:36   thank Mac Mini Colo net for providing [TS]

00:00:39   the bandwidth for this episode it's [TS]

00:00:42   where I host a couple Mac minis and you [TS]

00:00:43   can host yours there to check it out Mac [TS]

00:00:46   Mini : n slash 5x5 special deals just [TS]

00:00:49   free guns [TS]

00:00:51   it's Friday John siracusa Friday [TS]

00:00:55   I guess it is and you were on some kind [TS]

00:00:58   of work-related phone call I was are you [TS]

00:01:02   doing all right yeah you know my skype [TS]

00:01:05   mute button still doesn't work this is [TS]

00:01:07   really bothering me what is it you're [TS]

00:01:09   doing behind them the mute button all [TS]

00:01:11   the time [TS]

00:01:11   I don't know taking a drink clearing my [TS]

00:01:14   breath then I just like to be muted my [TS]

00:01:15   mother understood but since the mute [TS]

00:01:18   button works but there's no visual [TS]

00:01:19   indication I used to blank out the [TS]

00:01:21   little person with the sphere shaped [TS]

00:01:25   head yeah thing and they would like put [TS]

00:01:27   a line through it and dim it and you [TS]

00:01:28   could tell that it was muted now nothing [TS]

00:01:30   visually changes and the tooltip doesn't [TS]

00:01:32   even change to mute on mutes the only [TS]

00:01:34   way I can tell whether a mute is to go [TS]

00:01:36   to the menu are you on to date or you on [TS]

00:01:38   the newer version I haven't changed a [TS]

00:01:39   thing I'm on the old version I've always [TS]

00:01:41   been on the old version I never upgraded [TS]

00:01:42   to five and it's mystery to me why this [TS]

00:01:46   suddenly doesn't work maybe this is what [TS]

00:01:48   Skype does is that instead of just [TS]

00:01:51   completely breaking the older version to [TS]

00:01:53   get you to go to the new version they [TS]

00:01:55   just slowly eliminate features one by [TS]

00:01:57   one until eventually all you can do is [TS]

00:01:59   launch the app and it can't type can't [TS]

00:02:01   connect can't do anything then you'd be [TS]

00:02:02   forced to upgrade it's working because I [TS]

00:02:04   considered I'm like well this mute [TS]

00:02:06   doesn't work one of those I'm great how [TS]

00:02:07   bad could it be but I said now let me so [TS]

00:02:10   I'm sticking it out but if there's a [TS]

00:02:13   long period of time where you can't hear [TS]

00:02:14   me I'm probably muted and don't know it [TS]

00:02:16   alright alright you're a person [TS]

00:02:18   follow-up yes so in the last show we [TS]

00:02:21   talked about Mountain Lion a lot and I [TS]

00:02:24   was mentioning the way to one of the [TS]

00:02:27   ways to bypass gatekeeper is that you [TS]

00:02:29   could right click on an icon or [TS]

00:02:32   control-click whatever in the finder and [TS]

00:02:34   select open which nobody ever does but [TS]

00:02:36   which if you do will not prompt you to [TS]

00:02:40   say you sure you want to open this blah [TS]

00:02:42   blah blah and some of the chatroom [TS]

00:02:43   mentioned that it might be are you sure [TS]

00:02:45   it's not a bug and I said I wasn't sure [TS]

00:02:46   but then subsequently in the chatroom [TS]

00:02:48   TMC double underscore said that Apple [TS]

00:02:53   actually advertises this feature on its [TS]

00:02:55   site it's this is quoting from Apple's [TS]

00:02:57   website you can even temporarily [TS]

00:02:58   override your settings by ctrl-click by [TS]

00:03:00   ctrl-clicking [TS]

00:03:01   and install any app at any time [TS]

00:03:03   gatekeeper leaves all up to you so [TS]

00:03:04   they're advertising that feature it's [TS]

00:03:05   like power users don't be afraid all [TS]

00:03:08   this gatekeeper stuff is [TS]

00:03:09   basically not for you uh if you really [TS]

00:03:13   want to open something you'll never be [TS]

00:03:14   prevented from opening it just use this [TS]

00:03:16   thing that no one but you will ever do [TS]

00:03:18   so that's an interesting use of an [TS]

00:03:20   obscure feature that probably has gone [TS]

00:03:24   completely unused by anybody now [TS]

00:03:25   suddenly has a purpose but now it's not [TS]

00:03:27   a buck Marco was excited that he got to [TS]

00:03:32   correct something on the show I think I [TS]

00:03:35   mentioned the last show about how I was [TS]

00:03:38   hoping some new Mac pros would come out [TS]

00:03:39   but they were still waiting for those [TS]

00:03:41   and then I was upset about the Ivy [TS]

00:03:43   Bridge delay that had been reported [TS]

00:03:44   right first on Marco show he he talked [TS]

00:03:47   more about the ivory Bridget Lee and [TS]

00:03:48   apparently that story about the delay [TS]

00:03:49   was overblown and really only the [TS]

00:03:51   dual-core ones are delayed and Apple's [TS]

00:03:53   not going to use a dual-core ones [TS]

00:03:54   probably anyway so it's not that not [TS]

00:03:57   that important Apple but the second [TS]

00:04:00   thing is that he said that the Mac pros [TS]

00:04:01   are not waiting on Ivy Bridge they're [TS]

00:04:03   waiting on the xeon e5 Sandy Bridge CPUs [TS]

00:04:06   now of course that's also speculation [TS]

00:04:11   but it's probably I don't know I don't [TS]

00:04:13   even know if it's a good guess like it's [TS]

00:04:16   it you know the reason the Mac pros [TS]

00:04:18   aren't here is because you know I don't [TS]

00:04:19   know what the macros are waiting I says [TS]

00:04:20   we don't know what they're going to have [TS]

00:04:21   in them when they ship ah it could be [TS]

00:04:25   that Apple just entirely skips that [TS]

00:04:26   generation of CPU and doesn't really [TS]

00:04:28   smack pros for another year or something [TS]

00:04:29   so ah but his point is that it's [TS]

00:04:34   conceivable that Ivy Bridge delayed or [TS]

00:04:36   not could be completely immaterial to [TS]

00:04:38   the presence of Mac Pro's I just hope [TS]

00:04:41   they released something something with [TS]

00:04:42   the newer CPU than the ones they sell [TS]

00:04:44   now talked about Mac App Store only ap [TS]

00:04:50   is and mountain lion how you have to [TS]

00:04:52   sell your app to the Mac App Store if [TS]

00:04:53   you want to use certain api's and have [TS]

00:04:55   that the number of those api's is [TS]

00:04:57   getting is increasing in mountain lion [TS]

00:05:00   right and I was talking about the iCloud [TS]

00:05:03   api's which generally Mac App Store only [TS]

00:05:06   and I said there's no technical reason [TS]

00:05:09   why those API it couldn't be available [TS]

00:05:12   to non Mac App Store things and it's [TS]

00:05:14   kind of like a carrot and stick approach [TS]

00:05:15   where the carrot is you get to use these [TS]

00:05:17   funny pies and the stick is everybody [TS]

00:05:19   else will be using these cool api's and [TS]

00:05:21   you won't so you better get into the Mac [TS]

00:05:22   App Store [TS]

00:05:23   many many people suggested a more [TS]

00:05:27   reasonable reason why the iCloud api's [TS]

00:05:29   in particular are limited to mac app [TS]

00:05:31   store only applications and that was [TS]

00:05:35   that iCloud is a server that the [TS]

00:05:38   server-side service of iCloud costs [TS]

00:05:40   Apple money to run so letting any [TS]

00:05:43   developer use it is basically saying [TS]

00:05:45   Apple's going to run a bunch of servers [TS]

00:05:46   for you and you can use them and store [TS]

00:05:48   your data on them free of charge mm-hmm [TS]

00:05:51   because iCloud a3d customers as well [TS]

00:05:52   whereas if you sell your app to the Mac [TS]

00:05:54   App Store at the very least Apple gets a [TS]

00:05:56   30% cut of your sales and that will help [TS]

00:05:58   offset the server usage in iCloud I'm [TS]

00:06:01   not sure how much I buy that theory [TS]

00:06:03   first of all you're allowed to have free [TS]

00:06:05   apps in the NICAP store so that kind of [TS]

00:06:07   hurts you there I mean I guess they get [TS]

00:06:09   stuck at the $99 you for the developer [TS]

00:06:12   membership but you know but I mean I [TS]

00:06:16   don't know I I think that that's a good [TS]

00:06:19   reason that I should have mentioned but [TS]

00:06:20   I think if you were to ask Apple and got [TS]

00:06:24   them in a moment of honesty they would [TS]

00:06:25   not tell you that the reason the iCloud [TS]

00:06:28   API is are from Mac App Store only is [TS]

00:06:30   because they have to defer the cost [TS]

00:06:32   somehow because if that was the case if [TS]

00:06:34   they're really interested in deferring [TS]

00:06:35   the cost I don't think thirty percent of [TS]

00:06:37   app sales is going to come close to [TS]

00:06:39   deferring the cost of the iCloud server [TS]

00:06:42   infrastructure the cost of development [TS]

00:06:44   everything I think that Apple considers [TS]

00:06:46   iCloud something they have to build at [TS]

00:06:49   the fact that they're giving a lot away [TS]

00:06:50   for free to everyone shows that they [TS]

00:06:52   know they want everyone to use this much [TS]

00:06:54   more than they want to actually make [TS]

00:06:55   money off it so that's got to be a cost [TS]

00:06:57   Center for them maybe it helps that they [TS]

00:07:01   could possibly get a little bit of extra [TS]

00:07:03   money from Mac App Store sales to offset [TS]

00:07:05   the cost of iCloud but but I'm not sure [TS]

00:07:07   I think it's mostly because they're [TS]

00:07:09   trying to encourage people to develop [TS]

00:07:11   applications the way they want them to [TS]

00:07:13   in the way they want is you do it in the [TS]

00:07:14   Mac App Store your sandbox use that [TS]

00:07:16   cloud a POS and I guess they have some [TS]

00:07:20   degree of control over you as well if [TS]

00:07:21   you're a registered Mac developer versus [TS]

00:07:24   if you just download the download Xcode [TS]

00:07:27   for free and start hacking away at an [TS]

00:07:29   application and put up on your website [TS]

00:07:30   and hammers the iCloud servers or [TS]

00:07:32   something it's harder for them to come [TS]

00:07:34   and find you and say please stop doing [TS]

00:07:36   that [TS]

00:07:37   please stop disturbing your application [TS]

00:07:39   hammers are service so as always with [TS]

00:07:42   Apple lots of teary leaf reading and no [TS]

00:07:45   actual and so sorry that weird one there [TS]

00:07:48   is notifications and mountain lion the [TS]

00:07:49   big thing on the right side of the [TS]

00:07:50   screen the new icon in the upper right [TS]

00:07:52   corner that shows you notifications for [TS]

00:07:55   like ah you know you have a reminder set [TS]

00:07:57   or something in your calendar or a to-do [TS]

00:07:59   list item that has a reminder set its [TS]

00:08:01   whole bunch of ways you can be notified [TS]

00:08:02   notification center in mountain lion and [TS]

00:08:07   as a question of like is there any [TS]

00:08:09   server-side component that like what if [TS]

00:08:10   I don't even sign up for iCloud I still [TS]

00:08:12   have Notification Center why shouldn't [TS]

00:08:14   someone be able to write a Mac [TS]

00:08:15   application not distribute it through [TS]

00:08:17   the Mac App Store but still have their [TS]

00:08:20   application pop-up notifications like [TS]

00:08:22   say you're writing an IM client and you [TS]

00:08:23   want it to pop up a notification [TS]

00:08:24   notification center when a certain [TS]

00:08:25   person logs on because you've set up an [TS]

00:08:27   alert for that that has nothing to do [TS]

00:08:28   with iCloud that's not using any of [TS]

00:08:29   Apple's server resources it's purely a [TS]

00:08:31   completely client-side API to display a [TS]

00:08:34   notification in Notification Center and [TS]

00:08:37   I don't think Apple syncs those [TS]

00:08:38   notifications across your devices now [TS]

00:08:40   obviously if you're using an application [TS]

00:08:42   like I like iCal like calendar or [TS]

00:08:43   contacts or something that can be sync [TS]

00:08:45   to iCloud then there is a component but [TS]

00:08:47   it's kind of weird that Notification [TS]

00:08:49   Center which you can if you squint you [TS]

00:08:52   can look at and say that's an entirely [TS]

00:08:53   client-side new API a new piece of UI [TS]

00:08:55   that gets caught up into the Mac App [TS]

00:08:57   Store only thing so that also makes me [TS]

00:08:59   think that deferring server cost is not [TS]

00:09:02   the reason it is not the main reason why [TS]

00:09:05   they make an API a Mac App Store only [TS]

00:09:09   otherwise why would they make [TS]

00:09:10   notifications Mac App Store only someone [TS]

00:09:15   sent in a correction about how I'm [TS]

00:09:17   pronouncing speaking ill I am [TS]

00:09:18   applications however pronouncing the the [TS]

00:09:20   instant message application for the Mac [TS]

00:09:21   called Adi um how do you pronounce that [TS]

00:09:24   I've heard it two ways I've heard it a [TS]

00:09:27   diem and I've heard it a diem [TS]

00:09:29   I personally I say a diem but I have no [TS]

00:09:33   idea what if that's correct and I as has [TS]

00:09:36   been evidenced over the last several [TS]

00:09:38   years of doing these shows I'm the wrong [TS]

00:09:40   person to look to for correct [TS]

00:09:42   pronunciation of anything how do you say [TS]

00:09:44   it I say Adam - but I was corrected on [TS]

00:09:48   Twitter that it's supposed to be like [TS]

00:09:50   the words [TS]

00:09:50   Stadium and this is how the originator [TS]

00:09:53   their atomizer is er I guess he's the [TS]

00:09:58   guy originally wrote it he described as [TS]

00:10:00   the originator history it says it [TS]

00:10:01   arrives with stadium ATM stadium hmm [TS]

00:10:04   well you know this is something where [TS]

00:10:07   you have to the original creator has to [TS]

00:10:08   have some input on it I mean they're the [TS]

00:10:12   ones that came up with it they're the [TS]

00:10:14   ones that say now what's interesting is [TS]

00:10:16   you get the creator of Linux Linux [TS]

00:10:20   Torvalds says I pronounce Linux as Linux [TS]

00:10:24   that's how he says it I still in does he [TS]

00:10:27   say that I thought he wanted the other [TS]

00:10:29   way I line X ya know be the rhymes with [TS]

00:10:33   Linus no but his name is Lynn Asst in [TS]

00:10:36   his uh in his own but I've got it [TS]

00:10:39   backwards in my mind language yeah so uh [TS]

00:10:44   he is the creator of it but yet in his [TS]

00:10:50   very open-source mindset he says this is [TS]

00:10:53   how I pronounce it he doesn't say this [TS]

00:10:55   is how it must be pronounced so the [TS]

00:10:57   idiom guy is he saying this is the [TS]

00:11:00   correct way to pronounce it or is he [TS]

00:11:01   just saying this is this is how I say [TS]

00:11:03   this was a tweet from the adium a DM [TS]

00:11:07   Twitter account at twitter.com slash ad [TS]

00:11:10   I um I don't know who controls that [TS]

00:11:11   account but the acutest the cuida I'll [TS]

00:11:14   read it verbatim says the correct [TS]

00:11:15   pronunciation of Adi um is debatable [TS]

00:11:18   atomizer the originator called me the [TS]

00:11:22   stadium way so that's probably the [TS]

00:11:23   proper way so this is a personified [TS]

00:11:25   Twitter account we're saying call to me [TS]

00:11:26   I'm the I'm the application so ATM [TS]

00:11:29   sounds like the right way to do it all [TS]

00:11:31   right here I'm going to play this for [TS]

00:11:33   the benefit of the listeners now I have [TS]

00:11:35   the clip from Lin us himself let me see [TS]

00:11:38   how badly I got it wrong let us realign [TS]

00:11:41   us what's exactly your preferred [TS]

00:11:42   pronunciation um when I speak Swedish [TS]

00:11:45   its eNOS when I speak Finnish its he [TS]

00:11:48   knows when I speak English its Linus and [TS]

00:11:51   I really don't care how people browse my [TS]

00:11:53   name but Linux is always Linux well its [TS]

00:11:55   cuts off he says Linux is always Linux [TS]

00:11:57   so that's pretty unequivocal yeah so [TS]

00:12:01   that we just need the ATM guy to weigh [TS]

00:12:03   in [TS]

00:12:04   the lesson for this I think is a when [TS]

00:12:06   making a product and choosing a product [TS]

00:12:08   name and same thing for like domains and [TS]

00:12:12   stuff that there are some rules you can [TS]

00:12:14   think about like did you pick a good [TS]

00:12:16   domain name for your product or service [TS]

00:12:18   or application and you and you have to [TS]

00:12:21   come up with some criteria HOD how do I [TS]

00:12:22   what makes a domain name good what makes [TS]

00:12:25   a product name good and I would think if [TS]

00:12:26   you went through this exercise my big [TS]

00:12:29   things on domain names are easy to spell [TS]

00:12:33   if I tell you what it is not too long [TS]

00:12:35   not already taken not a synonym for [TS]

00:12:39   something that's bad you know it doesn't [TS]

00:12:43   have a lot of homonyms all sorts of [TS]

00:12:44   things like that and for a product name [TS]

00:12:46   this is similar like make it so that the [TS]

00:12:50   obvious pronunciation in whatever [TS]

00:12:52   language you consider most important is [TS]

00:12:55   the obvious pronunciation I think with [TS]

00:12:59   Linux he basically did that because [TS]

00:13:00   that's how everybody says that they CLI [TS]

00:13:02   em they're not they're not inclined to [TS]

00:13:04   say line-x [TS]

00:13:04   yeah but the you know the knowing the [TS]

00:13:09   creator's name was Linus got that [TS]

00:13:10   confused and the second thing is from [TS]

00:13:12   the very start have a pronunciation [TS]

00:13:14   guide I can't tell you how many sites I [TS]

00:13:17   go to for some open source tool or [TS]

00:13:19   whatever and I I would expect to see [TS]

00:13:22   like an FAQ or an about page or [TS]

00:13:24   something and I expect one of the items [TS]

00:13:26   perhaps the first time for that page to [TS]

00:13:28   be how do I pronounce your name just to [TS]

00:13:30   make sure like put that in there you [TS]

00:13:31   know and then there cross the people [TS]

00:13:33   like SQLite EE that product yeah it's [TS]

00:13:36   just it's hopeless and there's no [TS]

00:13:37   obvious way and it's just a big giant [TS]

00:13:39   mess but even something like my SQL you [TS]

00:13:43   can get in trouble real real fast if [TS]

00:13:45   your thing becomes popular and people [TS]

00:13:47   start pronouncing it one way and it's [TS]

00:13:49   not the way you want it so the moral of [TS]

00:13:52   story is pick pick names that have an [TS]

00:13:53   obvious pronunciation and documented [TS]

00:13:56   maybe documented unlike your very first [TS]

00:13:58   homepage will have a big banner with the [TS]

00:14:01   name of your product and right [TS]

00:14:02   underneath it those little pronunciation [TS]

00:14:03   things or a button you can click to hear [TS]

00:14:05   how it's pronounced [TS]

00:14:05   yeah alright on the last show I was [TS]

00:14:11   talking about and I was actually that [TS]

00:14:13   was the title to show the region of pain [TS]

00:14:15   where you have a ten point something [TS]

00:14:17   point-0 release that is probably full of [TS]

00:14:20   bugs and you have the point 1 and the [TS]

00:14:21   point 2 and it takes a while to crank up [TS]

00:14:23   with the with the you know minor [TS]

00:14:25   subversion until the thing kind of [TS]

00:14:27   settles down and all the bugs are ironed [TS]

00:14:29   out and I mentioned that other of Mac OS [TS]

00:14:32   10 releases that have had longer have [TS]

00:14:34   had a longer time in the market before [TS]

00:14:36   the successor came along got really high [TS]

00:14:38   up on that last number and I couldn't [TS]

00:14:40   remember one of them unto double digits [TS]

00:14:42   or something so TMC double underscore [TS]

00:14:43   again in the chatroom pointed me to a [TS]

00:14:46   link entitled a useless analysis of OS [TS]

00:14:50   10 release dates and then put that in [TS]

00:14:52   the show notes and that just shows how [TS]

00:14:54   far each one got so I'll just go through [TS]

00:14:56   the MIR Snow Leopard got 2.8 well we [TS]

00:14:58   don't know maybe it'll get the point out [TS]

00:14:59   but right now Snow Leopard is on point [TS]

00:15:00   eight leopard got the point eight Tiger [TS]

00:15:02   was the only one that won two double [TS]

00:15:04   digits Tiger out 2.11 panther got the [TS]

00:15:07   point nine jaguar got the point eight [TS]

00:15:09   puma got the point five and cheetah got [TS]

00:15:11   the point four so cheetah obviously was [TS]

00:15:13   like the big slow as molasses Harbor or [TS]

00:15:16   release we can't wait to get four things [TS]

00:15:18   this thing they ever play so like I [TS]

00:15:20   replace pretty quickly in Puma got [TS]

00:15:21   really plays really quickly too but all [TS]

00:15:22   the other ones got into like the you [TS]

00:15:24   know the eights nines or 11s so lion is [TS]

00:15:29   going to have to advance pretty quickly [TS]

00:15:30   if it ever expects to get into the you [TS]

00:15:33   know 8's 9s or 10s or 11s before ten [TS]

00:15:35   point eight lands this summer and I [TS]

00:15:37   don't think it will make it it may it [TS]

00:15:39   may end up ending its life kind of like [TS]

00:15:40   cheetah and Puma in the point four zero [TS]

00:15:43   point five which could possibly still be [TS]

00:15:46   in the region of pain alright and one [TS]

00:15:50   thing this one more thing this doesn't [TS]

00:15:51   really follow up but it's related to [TS]

00:15:53   iCloud and it's a story they went by my [TS]

00:15:56   eyes this week was that Apple's has [TS]

00:15:59   purchased some land in Oregon for [TS]

00:16:00   another data center did you see that [TS]

00:16:02   story oh yeah you know I didn't I didn't [TS]

00:16:07   read the whole thing about that so maybe [TS]

00:16:09   a where was it read and I'll get it into [TS]

00:16:11   the show notes it's in the show notes [TS]

00:16:12   already coming right now alright well [TS]

00:16:14   you said you were unprepared it turns [TS]

00:16:15   out you're prepared on my notes I want I [TS]

00:16:17   want you'll see I'll go off tracks I see [TS]

00:16:21   it okay yeah there it is right there [TS]

00:16:23   yeah it's not much to the story other [TS]

00:16:25   than the the title so this is in [TS]

00:16:27   Prineville Oregon they [TS]

00:16:29   about 160 acres for 5.6 million dollars [TS]

00:16:32   and the Apple spokesperson the only [TS]

00:16:35   statement quoted in this article that I [TS]

00:16:37   linked is the Apple spokesman says we [TS]

00:16:39   purchased the land and it's for a data [TS]

00:16:41   center so typical Apple disclosure what [TS]

00:16:44   more do you need to know I always [TS]

00:16:46   thought it was kind of weird with the [TS]

00:16:47   North Carolina data center how you know [TS]

00:16:49   Apple purchases big giant data center [TS]

00:16:51   it's gonna be really fancies costing [TS]

00:16:52   lots of money and everyone was all [TS]

00:16:53   excited about it and but it was on the [TS]

00:16:55   East Coast yeah and you're like well [TS]

00:16:58   don't you need like at least two data [TS]

00:16:59   centers if you're if you're interested [TS]

00:17:00   in covering the United States well don't [TS]

00:17:02   you need one on the west coast and one [TS]

00:17:03   on the East or maybe someone in the [TS]

00:17:04   middle I forget about worldwide for [TS]

00:17:06   anything like just having one data [TS]

00:17:08   center in the US is weird if you're if [TS]

00:17:10   you're the future of your company is [TS]

00:17:12   going to be like we're going to have an [TS]

00:17:13   online platform whatever like you know [TS]

00:17:15   because I'm sure Google and all the [TS]

00:17:17   other companies have data centers all [TS]

00:17:18   over the place they're just they have to [TS]

00:17:20   so this is kind of the other shoe [TS]

00:17:21   dropping at least in the u.s. to say [TS]

00:17:23   okay fine at least we'll have an East [TS]

00:17:24   Coast in a West Coast data center now a [TS]

00:17:26   hundred and hundred and sixty acres [TS]

00:17:28   acres are very confusing and misleading [TS]

00:17:31   for most people because an acre cent [TS]

00:17:34   seems like wow it's you know it's a [TS]

00:17:35   curse that's that's a huge space a [TS]

00:17:39   hundred and sixty acres correct me if [TS]

00:17:41   I'm wrong I think it's about a quarter [TS]

00:17:42   square mile I I can't do that math [TS]

00:17:46   Amanda lies how big an acre is in my [TS]

00:17:49   head but I don't know if 160 acres is [TS]

00:17:52   big or small for a data center I think [TS]

00:17:54   this it's that's a that's a decent sized [TS]

00:17:56   space but it's not like they just [TS]

00:17:58   acquired like a thousand square miles of [TS]

00:18:01   space this is this is a large place for [TS]

00:18:03   a nice-size building they'll have [TS]

00:18:05   parking for you know all of that stuff [TS]

00:18:07   but this is not like a campus yeah well [TS]

00:18:11   is it the other thing is I don't know [TS]

00:18:12   where all the apples data centers are [TS]

00:18:14   now we all know about the North Carolina [TS]

00:18:15   one because it was a big story and it [TS]

00:18:17   was you know that that that's the one [TS]

00:18:20   that it was gigantic and expensive and [TS]

00:18:22   it was they got a sweetheart deal on it [TS]

00:18:24   from the North Carolina people and there [TS]

00:18:25   was all sorts of stories related to that [TS]

00:18:27   but Apple surely has data centers all [TS]

00:18:29   over the place already and we don't know [TS]

00:18:31   where all those are but seeing Apple buy [TS]

00:18:34   another one I guess is a good sign for [TS]

00:18:36   the future of iCloud show they're [TS]

00:18:37   committed to it I still still wigs me [TS]

00:18:39   out and we had all show about this how [TS]

00:18:41   Apple is filling its data center [TS]

00:18:43   with hardware and a lot of software that [TS]

00:18:46   it does not right this gets back to the [TS]

00:18:48   Joel Spolsky thing which I think I [TS]

00:18:51   mentioned on the show when we discussed [TS]

00:18:52   this about don't outsource your core [TS]

00:18:54   competency and he actually discussed [TS]

00:18:55   things like this on on build and analyze [TS]

00:18:59   maybe also on back to work but on this [TS]

00:19:01   most recent build and analyze you're [TS]

00:19:03   talking with Marko about what is it that [TS]

00:19:05   you what is it that you do well and if [TS]

00:19:08   the answer is not I'm really good at [TS]

00:19:10   figuring out how small businesses should [TS]

00:19:12   pay taxes then maybe have someone else [TS]

00:19:14   do that so you can spend your time [TS]

00:19:16   developing software or whatever it is [TS]

00:19:17   that you consider the core competency of [TS]

00:19:19   your company so every company has a [TS]

00:19:20   whole bunch of things they have to do [TS]

00:19:22   and they have to decide which one of [TS]

00:19:23   these things is the thing that we're [TS]

00:19:26   good at and if you asked Apple they [TS]

00:19:28   would probably say like well the purpose [TS]

00:19:31   of our company is not to be really good [TS]

00:19:33   at managing payroll or to have really [TS]

00:19:35   good insurance on our facilities or to [TS]

00:19:38   have the you know to be really good at [TS]

00:19:41   cutting the lawns outside of our [TS]

00:19:42   buildings and have great parking you [TS]

00:19:44   know these are only things they do [TS]

00:19:45   things that they want to do but you say [TS]

00:19:47   is this what Apple is about and if the [TS]

00:19:49   answer is no then pay someone to cut the [TS]

00:19:51   lawn [TS]

00:19:51   perhaps you know why are some people to [TS]

00:19:53   figure out the facility stuff and you [TS]

00:19:55   know that stuff you can outsource but [TS]

00:19:57   you wouldn't for example outsource [TS]

00:20:00   designing the user interface for our [TS]

00:20:02   next version of iOS because Apple would [TS]

00:20:04   say that's what we do that's our that's [TS]

00:20:06   our core competency and we would never [TS]

00:20:08   outsource that well we're really [TS]

00:20:09   crunched we don't have enough people to [TS]

00:20:10   get iOS 6 out on time quick let's start [TS]

00:20:13   outsourcing some of the work to India [TS]

00:20:15   unless they're Apple's employees in [TS]

00:20:17   India they wouldn't like find an [TS]

00:20:18   outsourcing company and say hey we need [TS]

00:20:20   some screens designed for iOS 6 could [TS]

00:20:22   you take no because that's their core [TS]

00:20:24   competence they would never let that go [TS]

00:20:26   if Apple wants to base its future on [TS]

00:20:29   iCloud and have iCloud be the platform [TS]

00:20:30   for the next 10 years and all the other [TS]

00:20:32   things that they said many different [TS]

00:20:34   times don't you think they would have [TS]

00:20:36   more of a stake in what goes in those [TS]

00:20:39   data centers yes Apple software is going [TS]

00:20:41   in there too the all software that [TS]

00:20:42   implements iCloud is written by Apple [TS]

00:20:43   but they're not using Apple hardware [TS]

00:20:45   they're they're not using Apple like [TS]

00:20:47   infrastructure software and sort of [TS]

00:20:49   storage management or you know that [TS]

00:20:51   there's reportedly using Windows [TS]

00:20:54   and who knows what else they're using [TS]

00:20:57   like at some point it seems it will seem [TS]

00:21:00   weird maybe you know if Apple's strategy [TS]

00:21:04   to base their future on iCloud is [TS]

00:21:05   successful at some point in the future [TS]

00:21:07   many years from now will seem weird that [TS]

00:21:09   Apple's main and most important business [TS]

00:21:11   and infrastructure is running on [TS]

00:21:13   machines they don't build and running [TS]

00:21:16   infrastructure software that they didn't [TS]

00:21:18   make so they have to pay other companies [TS]

00:21:19   for that especially if that other [TS]

00:21:20   companies like Microsoft or Oracle or [TS]

00:21:22   whoever else it is versus the approach [TS]

00:21:25   of someone like Google who basically [TS]

00:21:27   does everything themselves because they [TS]

00:21:28   recognize this is our core competency we [TS]

00:21:30   run online services and we're not going [TS]

00:21:32   to be we're going to be masters of our [TS]

00:21:34   own destiny as far as this is desert [TS]

00:21:36   we're not going to be beholden to some [TS]

00:21:37   third party hardware software vendor to [TS]

00:21:39   deal with this we're going to do it all [TS]

00:21:40   ourselves it's kind of weird you say [TS]

00:21:43   well Apple doesn't make server operating [TS]

00:21:45   systems they don't make server hardware [TS]

00:21:46   they sort of did it but they try you [TS]

00:21:47   know does Apple have to make server [TS]

00:21:49   hardware maybe that's not their core [TS]

00:21:50   competency yeah [TS]

00:21:51   I think the hardware is easier to argue [TS]

00:21:53   that Apple doesn't necessarily happen [TS]

00:21:56   even though they do make computer [TS]

00:21:57   hardware they make consumer hardware not [TS]

00:21:58   Enterprise but the software like not [TS]

00:22:01   just the operating system but the [TS]

00:22:03   software that manages the data center [TS]

00:22:04   that that's the storage management [TS]

00:22:06   process management load balancing [TS]

00:22:09   failover just that whatever all the [TS]

00:22:11   things you have to do in the data center [TS]

00:22:13   if they're going to outsource all of [TS]

00:22:16   that that seems weird don't mean I don't [TS]

00:22:18   want to repeat the whole content of that [TS]

00:22:19   path that was a great show though but [TS]

00:22:22   that's still whenever I see these data [TS]

00:22:24   center stuff it wigs me out a little bit [TS]

00:22:26   autodidact in the chat room says 160 [TS]

00:22:30   acres is quarter square mile but that's [TS]

00:22:34   half a mile by half a mile so that [TS]

00:22:36   actually is a little bigger than I was [TS]

00:22:39   imagining well there you go there you go [TS]

00:22:43   all right now that's how big your plot [TS]

00:22:45   of land is about right alejandra my [TS]

00:22:47   parents house was on like a quarter acre [TS]

00:22:49   so that's how I visualize an acre just [TS]

00:22:50   take like the house that I grew up in [TS]

00:22:52   and I take the property and I multiply [TS]

00:22:55   by four you know what kind of a courage [TS]

00:22:57   do you have where you less than that [TS]

00:23:00   much less than barely enough the whole [TS]

00:23:01   house [TS]

00:23:06   but you're making it anyway yeah it's a [TS]

00:23:10   more densely populated here then I'm [TS]

00:23:12   Texas where everything's bigger so today [TS]

00:23:18   I thought we would talk about file [TS]

00:23:21   systems how do you feel about that I I [TS]

00:23:24   would love that [TS]

00:23:25   that'd be amazing people have been [TS]

00:23:27   really wanting to hear a ZFS discussion [TS]

00:23:31   for a long time why what is ZFS who [TS]

00:23:33   cares is it in OS 10 is it not an OS 10 [TS]

00:23:37   will it ever be in OS 10 is it dead is [TS]

00:23:38   alive [TS]

00:23:40   how far back are you going to take us [TS]

00:23:42   you're going to take us back to HFS hfs+ [TS]

00:23:44   this is where my notes become like my [TS]

00:23:48   problem notes is when I construct them I [TS]

00:23:50   tend to put too much into them and then [TS]

00:23:52   I realize oh for this one bullet point I [TS]

00:23:54   have five paragraphs of text which I [TS]

00:23:56   can't glance that and I don't want to [TS]

00:23:58   read verbatim like I need to condense it [TS]

00:24:01   so that I can look at like single line [TS]

00:24:03   bullet points in my outline view to say [TS]

00:24:05   okay remind remember to talk about that [TS]

00:24:07   remember to talk about that but instead [TS]

00:24:08   I have this big sprawling thing of text [TS]

00:24:10   site I've taken to putting certain [TS]

00:24:12   sections of the big sprawling wall of [TS]

00:24:14   text in bold so I can just glance it's [TS]

00:24:16   bold parts remind me so this may be a [TS]

00:24:18   little bit scattered but we'll see what [TS]

00:24:20   we can do here so the file system stuff [TS]

00:24:23   I thought I would start with what I [TS]

00:24:25   wrote about file systems in my Mac os10 [TS]

00:24:30   10.7 lion review and I think that was [TS]

00:24:33   the first time I'd really talked about [TS]

00:24:34   file systems in in a Mac OS tender you [TS]

00:24:36   had written about it on in separate [TS]

00:24:38   articles and about the possibility of [TS]

00:24:40   ZFS being a Mac it was 10 and what ZFS [TS]

00:24:42   was I remember the ZFS was first coming [TS]

00:24:44   out and I don't like 10 years ago or [TS]

00:24:46   something maybe eight years ago they had [TS]

00:24:48   all these cool videos online about them [TS]

00:24:50   and I would link them save it boy look [TS]

00:24:51   at this cool thing because I'm into file [TS]

00:24:52   systems but when it came time to do the [TS]

00:24:54   Mac OS 10 review I didn't have much to [TS]

00:24:55   say about it until there was a big rumor [TS]

00:24:57   about ZFS coming to Mac OS 10 and [TS]

00:25:01   basically what happened is that Sun went [TS]

00:25:03   to Apple and pitched them on their new [TS]

00:25:05   file system I said hey you know we've [TS]

00:25:07   got this new file systems called ZFS [TS]

00:25:08   maybe you've seen stuff on the web about [TS]

00:25:10   it let's we're going to come and pitch [TS]

00:25:12   your engineers so they did a [TS]

00:25:14   presentation and said here's the FS and [TS]

00:25:15   the engineers asked them hard questions [TS]

00:25:17   and so on and so forth [TS]

00:25:18   uh and the team or the people [TS]

00:25:22   responsible for the file system at Apple [TS]

00:25:25   sort of decided that yeah let's try the [TS]

00:25:27   ZFS think this could be our ticket to [TS]

00:25:29   having a new file system at Apple is not [TS]

00:25:31   above going outside for something for [TS]

00:25:35   some software that it needs the main [TS]

00:25:39   barrier that I can imagine to the [TS]

00:25:41   typical Apple mindset is well can we [TS]

00:25:43   control this so for example Apple [TS]

00:25:44   adopted the KHTML product project from [TS]

00:25:47   the KDE in the Linux platform and they [TS]

00:25:51   sort of made it their own by turning [TS]

00:25:52   into WebKit because the open-source [TS]

00:25:53   license allowed them to do that their [TS]

00:25:55   chant you know WebKit is still open [TS]

00:25:56   source it's an open source project but [TS]

00:25:57   Apple basically took over development of [TS]

00:26:00   that thing and gave it a new name and [TS]

00:26:03   went off and ran with it so even though [TS]

00:26:06   this is something created by someone [TS]

00:26:07   else they're not beholden to someone for [TS]

00:26:10   advances in their web browser technology [TS]

00:26:11   and similarly for the BSD layer and in [TS]

00:26:14   their operating system BSD is is free [TS]

00:26:17   for an open source and you know free for [TS]

00:26:19   commercial use as long as they release [TS]

00:26:21   the sources which they do is part of the [TS]

00:26:22   darwin open source releases and all that [TS]

00:26:24   stuff so the question was well ZFS can [TS]

00:26:27   we take this and run with it like we [TS]

00:26:29   like the fact that someone else made it [TS]

00:26:31   and will continue to develop it but if [TS]

00:26:32   we are ever find ourselves at [TS]

00:26:34   cross-purposes with sun will would be ok [TS]

00:26:36   so that was one of the concerns we like [TS]

00:26:39   the technology it seems like suns [TS]

00:26:42   open-source license allows us to do what [TS]

00:26:44   we did with WebKit or BSD or any of the [TS]

00:26:46   other open source things that we use [TS]

00:26:47   where we won't be under the thumb or [TS]

00:26:49   waiting on Sun to make certain changes [TS]

00:26:51   and you know we'll maintain our own port [TS]

00:26:54   and make sure that it works in Mac OS 10 [TS]

00:26:56   and go with that [TS]

00:26:57   and I'd remember how far we got I think [TS]

00:26:59   the farthest it got was like on Apple [TS]

00:27:01   calm web page for Snow Leopard it said [TS]

00:27:04   stuff about ZFS and you'll be able to [TS]

00:27:07   use the FS isn't that cool that was [TS]

00:27:09   early in the development process and [TS]

00:27:11   eventually that text disappeared from [TS]

00:27:12   apple.com and as we all know that Apple [TS]

00:27:16   did not ship their operating system with [TS]

00:27:18   CFS as the default file system or even [TS]

00:27:20   as a support of file system although you [TS]

00:27:22   could get it from Apple comm for a long [TS]

00:27:24   time you could download [TS]

00:27:24   here's apples port of ZFS and try it and [TS]

00:27:27   it was just buggy it not complete and [TS]

00:27:28   you're kind of glad they didn't do [TS]

00:27:29   anything with it it did they they might [TS]

00:27:31   have even really [TS]

00:27:32   a version of server with support for it [TS]

00:27:34   built into it but the thing is it [TS]

00:27:36   basically just fizzled and there's many [TS]

00:27:39   questions as to why what happened is it [TS]

00:27:42   because son was sued by NetApp over [TS]

00:27:45   patents in the file system or later [TS]

00:27:48   people say well it's because they knew [TS]

00:27:50   Oracle was going to buy son or because [TS]

00:27:52   Oracle did buy son or they decided that [TS]

00:27:55   ZFS was not good not a good technical [TS]

00:27:57   fit for them inside Apple I imagine [TS]

00:28:00   there were definitely two factions the [TS]

00:28:02   people who really thought ZFS was [TS]

00:28:03   awesome that Apple should adopt it and [TS]

00:28:04   the people who thought it's better for [TS]

00:28:07   server type platforms and not great for [TS]

00:28:09   a consumer platform but the upshot is [TS]

00:28:10   that here we are now in 2012 and the [TS]

00:28:14   defaults file system for Mac OS 10 is as [TS]

00:28:17   it ever was H of S Plus and H of s plus [TS]

00:28:20   is very old thousand so I thought during [TS]

00:28:21   the line review this would be a good [TS]

00:28:23   time for me to take stock of this and so [TS]

00:28:25   I did a section called the state of the [TS]

00:28:27   file system and in it I described a [TS]

00:28:31   little bit of the story and what what [TS]

00:28:34   the current file system is and why it [TS]

00:28:37   needs a change and I thought this was [TS]

00:28:39   something worth explaining because most [TS]

00:28:41   of the time you bring up file systems to [TS]

00:28:42   even-even to nerds but people who aren't [TS]

00:28:44   into file systems like you talk to [TS]

00:28:46   computer programs like Oh what do I care [TS]

00:28:48   what the file system is like can I get [TS]

00:28:49   files off of it [TS]

00:28:50   does it does it work then why does it [TS]

00:28:53   need to change and that's a reasonable [TS]

00:28:57   position again not just for regular [TS]

00:28:58   users regular users don't even know the [TS]

00:29:00   file system exists or what you're [TS]

00:29:01   talking about number for actual [TS]

00:29:02   programmers disabled HFS+ is fine like [TS]

00:29:04   it the thing is they don't have a way to [TS]

00:29:07   if you're not into files and technology [TS]

00:29:10   you don't have a good way to judge what [TS]

00:29:12   makes a good file system look at because [TS]

00:29:17   it basically you have reliability which [TS]

00:29:18   is well does it seem to work most of the [TS]

00:29:21   time and I guess you kind of have speed [TS]

00:29:25   but people tend to put speed onto the [TS]

00:29:27   hardware they're like oh this hard disk [TS]

00:29:28   is really slow or this operating system [TS]

00:29:30   is slow or this version of this [TS]

00:29:31   operating system is slower this [TS]

00:29:32   application is really slow to launch or [TS]

00:29:36   any of those type of things tend not to [TS]

00:29:37   be laid at the feet of the file system [TS]

00:29:39   because it's as constant and since it [TS]

00:29:41   hasn't changed and much in such a long [TS]

00:29:43   time it's hard [TS]

00:29:45   to say oh this would be faster if we had [TS]

00:29:47   a different file system or anything like [TS]

00:29:49   that [TS]

00:29:49   but there are better criteria for [TS]

00:29:52   judging a file system and file systems [TS]

00:29:53   have come a long way since hfs+ which is [TS]

00:29:56   derived from HFS and so on and so forth [TS]

00:29:58   so you could follow the link in the show [TS]

00:29:59   notes to that the link takes you [TS]

00:30:02   directly to that particular section of [TS]

00:30:03   my Lyon review and you can read about it [TS]

00:30:05   at length I'll summarize the few of the [TS]

00:30:10   things that I complained about first was [TS]

00:30:13   that [TS]

00:30:15   HMS Plus in my very long experience with [TS]

00:30:18   it tends to corrupt itself so it's not [TS]

00:30:24   not a hardware problem like a software [TS]

00:30:25   alright so you have you have disk disk [TS]

00:30:28   what's called disk first aid a long time [TS]

00:30:30   ago yeah it's called Disk Utility now I [TS]

00:30:32   if you run that on your Mac and say you [TS]

00:30:36   can you can verify a repair you can [TS]

00:30:38   verify your boot disk or you can repair [TS]

00:30:39   another disk so if you boot from a [TS]

00:30:41   different disc this is or even just hit [TS]

00:30:43   verify on your boot disk chances are [TS]

00:30:45   pretty good it will find problems and [TS]

00:30:47   what is it what is it has a bunch of [TS]

00:30:48   mumbo-jumbo it's going to say like in [TS]

00:30:50   correct extent block or bitmap error or [TS]

00:30:54   something you know it's going to say all [TS]

00:30:55   these weird stuff that you don't [TS]

00:30:56   understand [TS]

00:30:57   and they're going to show up a little [TS]

00:30:58   red text and you're going to be like [TS]

00:30:59   what the heck is that and then if it's [TS]

00:31:01   not your boot disk you can hit repair [TS]

00:31:02   and it will go through and grind grind [TS]

00:31:04   grind and find these little errors and [TS]

00:31:06   fix them and say repair successful then [TS]

00:31:08   you run verify again it'll say ok [TS]

00:31:09   everything's great this is not healthy [TS]

00:31:12   behavior for file systems where the file [TS]

00:31:16   system job is to keep track of where [TS]

00:31:17   everything is so when you ask our files [TS]

00:31:20   go say ok well where where is the data [TS]

00:31:22   for this file and if you don't think [TS]

00:31:24   about it too much like oh isn't it just [TS]

00:31:26   like in one big long line and it reads [TS]

00:31:28   off this now it's spread all over the [TS]

00:31:29   place and little pieces and it has to [TS]

00:31:30   keep track of where little pieces are [TS]

00:31:32   what order they go in you know and on [TS]

00:31:34   and on until like how many files are in [TS]

00:31:36   this directory where do I get to those [TS]

00:31:37   files what are the names of those files [TS]

00:31:39   all this information about where stuff [TS]

00:31:41   is on your disk the file system has data [TS]

00:31:44   structures that it manages to keep track [TS]

00:31:46   of that and of course those data [TS]

00:31:48   structures themselves are stored on the [TS]

00:31:50   disk and there's a whole big hierarchy [TS]

00:31:51   of layering for the storage but broadly [TS]

00:31:53   speaking there when I talk about file [TS]

00:31:56   system metadata it's the stuff that the [TS]

00:31:58   file system stores to know [TS]

00:31:59   where everything is and when it finds [TS]

00:32:00   these errors it make a lot of them maybe [TS]

00:32:02   seemingly benign like oh this should [TS]

00:32:05   this file was actually deleted but this [TS]

00:32:07   space used by this file still marked as [TS]

00:32:10   being in use in this thing that we use [TS]

00:32:11   to keep track of what pieces of disk are [TS]

00:32:13   used or not so I should just mark that [TS]

00:32:15   as not in use anymore [TS]

00:32:16   I don't know how that happened but I [TS]

00:32:18   should fix it all the way up to I can't [TS]

00:32:21   make heads or tails of this directory [TS]

00:32:23   entry I can't tell what's supposed to be [TS]

00:32:24   in this directory or the directory entry [TS]

00:32:27   says it's supposed to be 50 files in [TS]

00:32:28   here but I can only see two so [TS]

00:32:29   something's messed up here so I'm not [TS]

00:32:31   quite sure what to do should I change it [TS]

00:32:33   so that the number says two instead of [TS]

00:32:35   50 or maybe they really work 50 files in [TS]

00:32:37   here and I should go try and scan the [TS]

00:32:38   disk and see if I can reconstruct where [TS]

00:32:39   those 50 files are all this again is not [TS]

00:32:41   Hardware a problem your disk is not [TS]

00:32:43   failing although this can be part of it [TS]

00:32:45   but these things can happen and do have [TS]

00:32:47   a non hfs+ when your hardware is [TS]

00:32:50   perfectly fine and this is not this is [TS]

00:32:52   not a good thing not a good thing at all [TS]

00:32:54   it's very troubling would you say yes [TS]

00:32:57   so Apple has been trying to address this [TS]

00:32:59   with their file system for a while the [TS]

00:33:01   HMS plus was derived from HFS which was [TS]

00:33:03   developed in like 1986 or something so [TS]

00:33:05   it's really old and file system [TS]

00:33:07   technology has been marching on and [TS]

00:33:09   especially in the server space you want [TS]

00:33:11   your servers disk to be resilient in the [TS]

00:33:14   case of you know that someone yanked out [TS]

00:33:16   the power club plugged in the data [TS]

00:33:18   center or you got a kernel panic or [TS]

00:33:19   something like that you don't want to [TS]

00:33:20   lose all your data in the server space [TS]

00:33:21   so that was the first the server was the [TS]

00:33:23   first realm of let's care about our data [TS]

00:33:26   and let's not accept that off some will [TS]

00:33:29   pull the plug well tough luck [TS]

00:33:30   everything's gone or you might lose [TS]

00:33:31   everything [TS]

00:33:33   so journaling is one of the first things [TS]

00:33:36   they use to address this where the [TS]

00:33:37   filesystem writes a little or writes a [TS]

00:33:39   little diary of like okay I'm going to [TS]

00:33:41   do this and then it does it and then it [TS]

00:33:42   goes back to the list and says okay I [TS]

00:33:44   did that thing that's journaling in a [TS]

00:33:46   broad sense and what this lets you do is [TS]

00:33:49   if the plug goes out in the middle of [TS]

00:33:53   you doing something when you go back on [TS]

00:33:54   you should have a little entry in the [TS]

00:33:55   journal that says okay I'm going to do [TS]

00:33:57   this and then you can look on your disk [TS]

00:33:58   and see that it's like half done and you [TS]

00:34:00   if you have enough information you can [TS]

00:34:02   you can finish doing at that point [TS]

00:34:03   because you have your little journal [TS]

00:34:04   entry of what you're going to do or you [TS]

00:34:06   can say let's go back to the time before [TS]

00:34:08   that happened [TS]

00:34:09   and pretend that didn't happen at all so [TS]

00:34:11   it's trying to keep track of what you [TS]

00:34:14   plan to do and whether it's been done or [TS]

00:34:15   not so when you when you start up an [TS]

00:34:18   HFS+ volume so they add a journaling to [TS]

00:34:21   hfs+ that it had some that a term it [TS]

00:34:25   terminated abruptly during the middle of [TS]

00:34:27   something it will say replaying the [TS]

00:34:28   journal so it will see those little [TS]

00:34:29   journal entries and replay them and try [TS]

00:34:31   to get you back into a consistent state [TS]

00:34:32   that's what they're always looking for [TS]

00:34:34   in file systems is things have to be in [TS]

00:34:36   a consistent state all the little blocks [TS]

00:34:39   of data have to be accounted for every [TS]

00:34:40   files little blocks have to be there all [TS]

00:34:42   the metadata has to match up so like I [TS]

00:34:44   said if I say that there are two files [TS]

00:34:46   in this directory [TS]

00:34:47   there should be two files in there when [TS]

00:34:48   I look at my little list of files in the [TS]

00:34:49   directory inside the file system and I [TS]

00:34:51   see five but the directory says I'm only [TS]

00:34:53   supposed to have three something's wrong [TS]

00:34:54   and I don't quite know what it is now [TS]

00:34:57   this doesn't happen like crazy and hfs+ [TS]

00:35:01   disk can take a huge number of these [TS]

00:35:03   little tiny errors and you won't notice [TS]

00:35:05   them that's why I think you know [TS]

00:35:07   everybody after they listen to the show [TS]

00:35:08   could should reboot from their boot CD [TS]

00:35:11   that came with their Mac or reboot from [TS]

00:35:14   an alternate drive or from the recovery [TS]

00:35:16   drive they're using a line to hold down [TS]

00:35:17   command are when you boot and run Disk [TS]

00:35:19   Utility on your boot disk and click the [TS]

00:35:22   repair button or just click verify and [TS]

00:35:25   see if you find any errors what you'd [TS]

00:35:27   want to find is zero errors because [TS]

00:35:30   that's that shows that your software is [TS]

00:35:32   resilient to the kinds of weird failures [TS]

00:35:34   that you might have and the worst thing [TS]

00:35:36   about HIV+ I think is that you will get [TS]

00:35:39   errors even though as far as you're [TS]

00:35:42   concerned nothing bad happened you [TS]

00:35:43   didn't there was no power failure [TS]

00:35:45   no one kicked out the plug nothing like [TS]

00:35:47   that see you're like well shouldn't you [TS]

00:35:49   know under normal even if applications [TS]

00:35:51   crash under normal operation like I [TS]

00:35:54   don't care if the application crashes if [TS]

00:35:56   the application crashes when it's in the [TS]

00:35:57   middle of writing a file it should be up [TS]

00:35:58   to the file system you know that the i/o [TS]

00:36:01   interface isn't its job to at least make [TS]

00:36:03   sure the data is consistent yet your [TS]

00:36:04   your file itself could be scrambled the [TS]

00:36:06   contents of the file could be scrambled [TS]

00:36:08   but the file system should not oh yeah [TS]

00:36:09   those is file it's got 20 bytes in it [TS]

00:36:11   those 20 bytes are garbage because the [TS]

00:36:13   app crashed and just spit out random [TS]

00:36:14   memory into it whatever but it's exactly [TS]

00:36:16   20 bytes I know how big it is I'm not [TS]

00:36:18   confused about where this file is I'm [TS]

00:36:20   not confused about the the blocks that [TS]

00:36:22   are allocated to a Miskin nothing [TS]

00:36:23   everything is perfectly consistent as [TS]

00:36:24   far as the file system goes those are [TS]

00:36:26   the kind of errors that Disk Utility is [TS]

00:36:28   finding it's finding inconsistencies in [TS]

00:36:30   keeping track of your data on disk [TS]

00:36:32   doesn't care what the date is they could [TS]

00:36:34   all be garbage so file system is not [TS]

00:36:35   concerned about that it's just supposed [TS]

00:36:36   to keep track of everything so that [TS]

00:36:40   bothers me and the root cause of you [TS]

00:36:43   know why that happens are there many [TS]

00:36:45   different reasons why the happens we'll [TS]

00:36:46   get into a little bit later but the fact [TS]

00:36:48   that does happen is bad the second one [TS]

00:36:51   is some issue that almost no one seems [TS]

00:36:55   to care about except for CFS nerds and I [TS]

00:36:57   think many many more people should care [TS]

00:36:59   about and I've been excited as one of [TS]

00:37:01   the reason I was so excited about ZFS is [TS]

00:37:03   data integrity this is getting it to [TS]

00:37:06   hardware errors where you write a [TS]

00:37:09   program that writes out the numbers 1 2 [TS]

00:37:10   3 4 5 - to a file on disk you want to be [TS]

00:37:14   able to come back 6 months from now and [TS]

00:37:15   read that file and get 1 2 3 Pro 5 back [TS]

00:37:17   ah and one of the reasons you might not [TS]

00:37:21   get your contents back is a hardware [TS]

00:37:23   problem because remember these you know [TS]

00:37:25   hard disks especially we're magnetizing [TS]

00:37:28   little tiny pieces of something on a [TS]

00:37:30   spinning platter and you know it's [TS]

00:37:33   possible for one of those bits to flip [TS]

00:37:34   the other way cosmic rays coming from [TS]

00:37:37   the sky hitting your flash chips all [TS]

00:37:38   sorts of hardware related reasons that [TS]

00:37:40   are not the fault of the software that [TS]

00:37:41   Apple writes that could make the data [TS]

00:37:45   that you previously wrote to disk be [TS]

00:37:47   incorrect so this is mostly esoteric [TS]

00:37:52   academic type of thing or like oh well [TS]

00:37:54   you know cosmic rays and bits flipping [TS]

00:37:56   that happens so rarely and who really [TS]

00:37:57   cares and if one bit flips and your JPEG [TS]

00:37:59   image you know who really cares doesn't [TS]

00:38:01   make a difference but what I quoted in [TS]

00:38:04   my line reviewer are a reference to some [TS]

00:38:07   studies about ok well how often does [TS]

00:38:08   this happen so say it happens like you [TS]

00:38:11   know one in every million bits flips [TS]

00:38:14   like once a year or something like [TS]

00:38:15   Allah's no problem and when most of [TS]

00:38:18   these file systems were designed disks [TS]

00:38:20   were tiny they were like you know the [TS]

00:38:21   biggest hard disk you could get was 10 [TS]

00:38:23   megabytes that's not ten gigabytes [TS]

00:38:24   people is 10 megabytes you know that's [TS]

00:38:27   that was a big hunk and hard drive that [TS]

00:38:29   you buy for $1000 and so I said yeah we [TS]

00:38:31   get one bit flip error and every you [TS]

00:38:34   know 100 gigabytes of data but you know [TS]

00:38:36   but our drives are ten megabytes that we [TS]

00:38:38   would have to let that drive sit there [TS]

00:38:39   for 300 years before one of those bit [TS]

00:38:41   slipped so fast-forward to today and [TS]

00:38:46   it's not unheard of to have a terabyte [TS]

00:38:49   hard drive so all of a sudden that one [TS]

00:38:50   it's like the I've said this before and [TS]

00:38:52   mangled it and I'll do it again the one [TS]

00:38:54   in a million chance that means you know [TS]

00:38:55   a hundred people in China could do it or [TS]

00:38:57   a thousand people in China could do it I [TS]

00:38:58   still can't to him but the point is once [TS]

00:39:00   you get two really big numbers [TS]

00:39:01   that little info intestinal odds of data [TS]

00:39:05   of you know hardware-based the bit rot [TS]

00:39:08   suddenly become significant so that I [TS]

00:39:11   did I get this quote see this is where [TS]

00:39:13   my notes are falling down here's the [TS]

00:39:15   quote from the study from from 2010 they [TS]

00:39:18   studied one and a half million drives [TS]

00:39:20   over a course of 41 months and they [TS]

00:39:23   showed that 400,000 blocks had errors 8% [TS]

00:39:27   of which were discovered during raid [TS]

00:39:28   reconstruction we'll talk about rate in [TS]

00:39:30   a little bit so when you say that when [TS]

00:39:34   you do the odds they look you know like [TS]

00:39:37   oh I'm probably safe but when you think [TS]

00:39:38   about the idea that 400,000 blocks have [TS]

00:39:42   gone bad [TS]

00:39:43   across 1.5 million disks but that's not [TS]

00:39:47   great odds is like oh you know what is [TS]

00:39:49   that a one-in-four error chance that you [TS]

00:39:52   have a bad block on your disk those I [TS]

00:39:55   don't like those odds [TS]

00:39:56   because blocks are actually you know [TS]

00:39:58   pretty big and so in a certain point it [TS]

00:40:02   becomes someone's responsibility to make [TS]

00:40:04   sure that the data on disk is correct [TS]

00:40:07   and we just keep getting more and more [TS]

00:40:10   data and I think the point that I made [TS]

00:40:12   in line article and if not I'll make it [TS]

00:40:14   now is that the data we're putting the [TS]

00:40:16   digitally storing is becoming more and [TS]

00:40:19   more important like if you lost your [TS]

00:40:20   applications you can redownload them and [TS]

00:40:22   if you lost your files well hopefully [TS]

00:40:24   you had a backup of Monopoly those [TS]

00:40:25   backups aren't corrupted but we all have [TS]

00:40:28   our photos on our computers and our [TS]

00:40:30   photos of our kids instead of most of us [TS]

00:40:31   don't print them and put them in albums [TS]

00:40:32   anymore yeah if you lost all the photos [TS]

00:40:34   of your kids it's a little bit more [TS]

00:40:36   upsetting than even losing like your [TS]

00:40:39   novel in progress or some work thing [TS]

00:40:41   that you did our videos of your kids or [TS]

00:40:44   any you know the stuff we put in our [TS]

00:40:45   computers is becoming extremely [TS]

00:40:47   important and to lose it because [TS]

00:40:50   as well you know hardware is imperfect [TS]

00:40:51   and I had had a huge you know I have a [TS]

00:40:54   12 megapixel camera which you can you [TS]

00:40:56   know which aren't that expensive these [TS]

00:40:57   days and I took tons of pictures of my [TS]

00:40:59   kid and I filled up my terabyte hard [TS]

00:41:00   drive with tons of images and I had [TS]

00:41:03   backups and everything but the backups [TS]

00:41:05   are corrupted because the main disk was [TS]

00:41:06   corrupted and I think it's the [TS]

00:41:09   responsibility of the hardware and [TS]

00:41:12   software vendor to eventually address [TS]

00:41:14   this and say we need a way to know to [TS]

00:41:16   know that the data is correct and so the [TS]

00:41:18   they obvious way to do that is to do [TS]

00:41:20   check sums on all your data which is you [TS]

00:41:21   write the data to the disk and then you [TS]

00:41:23   write something derive from that data so [TS]

00:41:25   if that data ever changes when you read [TS]

00:41:27   arrive that thing from it it will not [TS]

00:41:29   match this doesn't fix the problem of [TS]

00:41:32   the bits flipping that bits are always [TS]

00:41:33   going to be bad but what you want to do [TS]

00:41:34   is you want to know when it's bad so for [TS]

00:41:38   example your backup program could refuse [TS]

00:41:40   to backup that new version of that file [TS]

00:41:43   and say well I've got an old backup [TS]

00:41:45   version of that file but the new version [TS]

00:41:47   something's happened to it on the disk [TS]

00:41:48   and it's messed up because the checksum [TS]

00:41:50   ship says it should should be this but [TS]

00:41:52   when I read the actual data off that [TS]

00:41:53   block I get something different so that [TS]

00:41:55   data is messed up now and I don't want [TS]

00:41:56   to copy that on top of your old version [TS]

00:41:58   that's what happens with corruptions if [TS]

00:42:01   you don't know about it eventually all [TS]

00:42:03   your backups are filled with that [TS]

00:42:04   corruption - like you you literally have [TS]

00:42:06   no way to know if you were to go to an [TS]

00:42:08   HFS+ disk now and say oh can I check [TS]

00:42:10   this disk to make sure all my pictures [TS]

00:42:12   and stuff are correct no there's no way [TS]

00:42:13   you can check because you don't know [TS]

00:42:14   what the date is supposed to be you can [TS]

00:42:16   read what it is in sight well like [TS]

00:42:18   here's the contents of this JPEG is that [TS]

00:42:20   right maybe you could check like is it [TS]

00:42:22   uh is it does it comply with the JPEG [TS]

00:42:23   format but inevitably there's gonna be [TS]

00:42:25   big sections of data on your disk and [TS]

00:42:27   you just have to go I guess I don't know [TS]

00:42:30   is that what we wrote originally [TS]

00:42:32   probably commit you open it up in the [TS]

00:42:34   JPEG view and you can look at it if you [TS]

00:42:36   see like a big squiggly lines or this [TS]

00:42:37   big you know colored rainbow stripes on [TS]

00:42:41   it you know it's messed up but that you [TS]

00:42:44   have no way to really know and [TS]

00:42:45   executable programs are even worse [TS]

00:42:46   because you have to like run it and make [TS]

00:42:48   sure you you execute the piece of code [TS]

00:42:49   that's on that block it's it's it's [TS]

00:42:52   scary to think that you have you know [TS]

00:42:54   millions and billions and billions of [TS]

00:42:56   bytes of data and you have no idea if [TS]

00:42:59   any of it is correct so ZFS [TS]

00:43:03   one of the many things the FS did was [TS]

00:43:04   address that and they said we were going [TS]

00:43:06   to have end-to-end data integrity that [TS]

00:43:09   we will just put check sums at [TS]

00:43:11   everything and so anything is screwed up [TS]

00:43:13   anywhere we will be able to find out and [TS]

00:43:16   in the course of development of ZFS they [TS]

00:43:18   did interesting things like found bugs [TS]

00:43:20   in discs firmware or our drivers or [TS]

00:43:23   other sorts of pieces of the storage [TS]

00:43:26   stack that people always assumed were [TS]

00:43:28   infallible like you know I have some [TS]

00:43:30   card that has a bunch of disks attached [TS]

00:43:33   to the stick into a server and ZFS would [TS]

00:43:35   start complaining like all your file [TS]

00:43:36   system is broken keeps complaining about [TS]

00:43:38   errors there's no errors here and [TS]

00:43:39   eventually you find out it's a bug in [TS]

00:43:41   some chip on the controller for that [TS]

00:43:42   disk drive and you never would have [TS]

00:43:44   found that before it just would have [TS]

00:43:45   been silently corrupting your data in [TS]

00:43:46   and you would have no idea that it was [TS]

00:43:48   doing and the final thing that ZFS [TS]

00:43:53   brings a table that hfs+ needs just [TS]

00:43:56   really this gets into the bait of what [TS]

00:43:58   is a file system and what should it do [TS]

00:44:00   but this is a concept of thing called [TS]

00:44:01   logical volume management and the quick [TS]

00:44:03   summary of that is the idea that a [TS]

00:44:06   single icon and your desktop it looks [TS]

00:44:07   like one container of first stuff could [TS]

00:44:10   be made up of more than one physical [TS]

00:44:13   thing it's divorcing the physical from [TS]

00:44:16   the logical so physical volumes are like [TS]

00:44:18   a big square metal thing that's a hard [TS]

00:44:21   disk or a bunch of chips that make up an [TS]

00:44:22   SSD and a logical volume could be is a [TS]

00:44:27   concept that says ok take a little bit [TS]

00:44:29   storage from here a little bit storage [TS]

00:44:30   from there a little storage from there [TS]

00:44:31   and I pull it all together into this one [TS]

00:44:32   big thing and this is what it looks like [TS]

00:44:34   as far as the operating system is [TS]

00:44:35   concerned but behind the scenes the [TS]

00:44:36   physical reality is different and [TS]

00:44:37   there's all sorts of ways you can mix [TS]

00:44:38   and match that many people will say that [TS]

00:44:42   logical volume management supposed to be [TS]

00:44:43   separate from the file system the file [TS]

00:44:45   system should just be the concern of the [TS]

00:44:46   you know physical device and then [TS]

00:44:48   logical volume management's layer above [TS]

00:44:49   that ZFS sort of combines it all into [TS]

00:44:51   one big thing now getting to I'm [TS]

00:44:58   creeping up on something believe it or [TS]

00:44:59   not we should we should creep up on our [TS]

00:45:01   our first sponsor that's a good idea [TS]

00:45:04   all right you can take a breather and [TS]

00:45:06   use try try your mute switch all right [TS]

00:45:09   here we go [TS]

00:45:09   all right first sponsor its apps fire [TS]

00:45:14   these guys are really cool I've been [TS]

00:45:16   used [TS]

00:45:16   this site a lot since then in a week or [TS]

00:45:19   two before they wanted to sponsor so I'm [TS]

00:45:22   pretty impressed with what they're doing [TS]

00:45:23   and here basically what's the problem [TS]

00:45:26   what are they trying to solve I'll tell [TS]

00:45:27   you finding your way around the App [TS]

00:45:28   Store it's it's a big challenge I mean [TS]

00:45:30   you've been you've been reading the news [TS]

00:45:32   maybe over the last week Apple [TS]

00:45:33   apparently acquired some company because [TS]

00:45:35   it admittedly they don't come out and [TS]

00:45:37   say it's a problem but when they're [TS]

00:45:39   acquiring a company that does things [TS]

00:45:41   like this to try and make the experience [TS]

00:45:43   better that's that's an admission that [TS]

00:45:45   they want to make it better well you [TS]

00:45:47   don't have to wait for them to figure [TS]

00:45:48   out how to do that you've got apps fire [TS]

00:45:50   today and this is because finding your [TS]

00:45:53   way around the App Store figuring out [TS]

00:45:55   what apps to use it can be it can be a [TS]

00:45:57   big challenge and that's the challenge [TS]

00:45:59   that people are faced with it's even a [TS]

00:46:00   tougher challenge if you are the creator [TS]

00:46:03   of an app and you want to get your app [TS]

00:46:04   there and you want to get it out in [TS]

00:46:05   front of people you want to make it easy [TS]

00:46:07   for them to find forget using I mean 4k [TS]

00:46:10   using iTunes in the App Store that [TS]

00:46:11   that's that's tough this is where apps [TS]

00:46:13   fire comes into the picture it's a [TS]

00:46:14   discovery and promotion service [TS]

00:46:16   specifically for mobile apps they create [TS]

00:46:20   a very slick guide popular guides to the [TS]

00:46:23   App Store and to Android users it's used [TS]

00:46:26   by millions of users around the world [TS]

00:46:27   you can discover the App Store in a [TS]

00:46:30   whole new way the apps of your friends [TS]

00:46:32   the apps of text celebrities absolute [TS]

00:46:34   that drop in price all of this stuff [TS]

00:46:36   they track it all at they make it very [TS]

00:46:39   easy for you as a creator to put your [TS]

00:46:42   app in front of a really really good and [TS]

00:46:44   growing audience they don't pay users [TS]

00:46:48   there's no scam involved they're not [TS]

00:46:50   paying people to download it users [TS]

00:46:52   download it if they want it's that [TS]

00:46:53   simple [TS]

00:46:54   it's just a way to get better quality [TS]

00:46:56   users faster so you can get 10% off of [TS]

00:47:00   this service if you use the code five by [TS]

00:47:02   five and apparently they're giving away [TS]

00:47:05   some free ad campaigns so so reach out [TS]

00:47:07   get in touch you go to apps feiyr.com [TS]

00:47:08   slash 5x5 you can even contact them [TS]

00:47:11   directly they set up a special email [TS]

00:47:13   address five by five at apps feiyr.com [TS]

00:47:16   so please do go check those guys out [TS]

00:47:18   very cool service all right I'm going to [TS]

00:47:24   talk a little bit about raid raid now [TS]

00:47:26   first of all we should come right out [TS]

00:47:28   and say raid [TS]

00:47:30   is a backup and all you need to do is [TS]

00:47:32   have your files on a raid system and [TS]

00:47:34   your fully backed up nothing to worry [TS]

00:47:36   about [TS]

00:47:36   sit back there's no that folks that's [TS]

00:47:39   sarcasm wait wait [TS]

00:47:42   raid isn't backup john rate is not a [TS]

00:47:45   back oh come on the people in about who [TS]

00:47:48   know what ray did so yeah redundant [TS]

00:47:50   array of independent it is inexpensive [TS]

00:47:54   and expensive yeah that they changed it [TS]

00:47:56   didn't they change it wonder they [TS]

00:47:58   probably did it because they realized [TS]

00:47:59   that they're making tons of money [TS]

00:48:00   selling $3,000 hard drive but yes [TS]

00:48:03   originally it was in it was inexpensive [TS]

00:48:05   disks the concept is you could take a [TS]

00:48:07   bunch of these cheap discs throw them [TS]

00:48:10   together and you could merge them into [TS]

00:48:12   one big volume and that volume might [TS]

00:48:15   maybe behind the scenes it was just [TS]

00:48:17   striped to make everything a lot faster [TS]

00:48:19   maybe a stripe for redundancy maybe it [TS]

00:48:20   was mirroring going on you could have [TS]

00:48:22   one drive found the others would keep [TS]

00:48:23   working and it was the holy grail of [TS]

00:48:26   storage it still maybe is well no it [TS]

00:48:30   definitely isn't anymore nothing like [TS]

00:48:32   originally it was one of the first [TS]

00:48:33   efforts to formalize the idea of I have [TS]

00:48:35   so much stuff that either it can't fit [TS]

00:48:37   on one disk at the current disk sizes or [TS]

00:48:39   I want more performance you know it's a [TS]

00:48:42   way to scale storage say look well I've [TS]

00:48:44   got one hard drive and it's this fast [TS]

00:48:46   can I get two hard drives and have it be [TS]

00:48:48   twice as fast well only if you're [TS]

00:48:50   reading one thing from this one one [TS]

00:48:51   thing from that one it's like well I [TS]

00:48:52   want I want scalable storage I want to [TS]

00:48:54   be able to add drives and either get [TS]

00:48:55   more performance and more redundancy or [TS]

00:48:57   both you know if it and then you get [TS]

00:48:59   into situation like where okay so I'm [TS]

00:49:01   putting my data on on these five disks [TS]

00:49:03   or six disks or whatever if one of the [TS]

00:49:05   disks fails do I lose all my data well [TS]

00:49:07   no that's going to the raid levels which [TS]

00:49:09   I'm not going to get into it's archaic [TS]

00:49:10   and no one should really need to know [TS]

00:49:11   raid levels but there are many different [TS]

00:49:14   schemes like well if you just merely [TS]

00:49:16   spread the data in five separate pieces [TS]

00:49:17   across these drives or whatever yeah if [TS]

00:49:19   you lose one disk everything is gone [TS]

00:49:20   because one fifth of your data is gone [TS]

00:49:21   and in four-fifths of data is no good [TS]

00:49:24   deal so that would but do you get better [TS]

00:49:26   performance out of that and then you [TS]

00:49:27   could go to mirroring where I'll get [TS]

00:49:29   like two drives and everything is on [TS]

00:49:30   both drives so if one drive fails [TS]

00:49:32   obviously I've still got all my data [TS]

00:49:33   because it was completely copied alright [TS]

00:49:35   and there are other schemes where they [TS]

00:49:36   will spread the data across of a couple [TS]

00:49:39   of drives and use one other drive to [TS]

00:49:41   write something that's to derive from [TS]

00:49:42   all that [TS]

00:49:43   it is such that if you lose one of the [TS]

00:49:44   drives you can read arrived what must [TS]

00:49:47   have been on that drive from the exist [TS]

00:49:49   the existing set of drives it's there so [TS]

00:49:52   now all of a sudden you have a situation [TS]

00:49:54   where you you're spreading data across [TS]

00:49:56   multiple drives you don't completely [TS]

00:49:58   duplicate the drives like you don't need [TS]

00:50:00   to have double the number drives you had [TS]

00:50:01   to have before and if you lose one disk [TS]

00:50:03   you can rebuild what the contents of [TS]

00:50:04   that disk were as you notice none of [TS]

00:50:07   these things address in any way the data [TS]

00:50:11   integrity issue like they just assumed [TS]

00:50:13   the disks when you write the data to [TS]

00:50:15   this when you read it back later that [TS]

00:50:16   will be the did they they do have the [TS]

00:50:19   parity thing there but I'll get to that [TS]

00:50:21   in a second so I have a link in the show [TS]

00:50:23   notes to Jeff bond Wix blog Jeff bond [TS]

00:50:25   wick is the was the leader of the team [TS]

00:50:28   that developed ZFS at Sun what is this [TS]

00:50:31   thing he was a vice president son and [TS]

00:50:33   senior Software Architect as he led the [TS]

00:50:35   team that developed ZFS for Solaris so I [TS]

00:50:41   think what the title of his blog post [TS]

00:50:42   this but his thing he talks about why [TS]

00:50:45   why ZFS has this thing in as part of its [TS]

00:50:50   product where you can where you can put [TS]

00:50:53   data on more than one disk isn't that [TS]

00:50:54   the the domain of the of logical volume [TS]

00:50:57   and you're like isn't that a layer above [TS]

00:50:59   the filesystem the fastest and just be [TS]

00:51:00   concerned with what it does and it kind [TS]

00:51:02   of goes through like why they develop [TS]

00:51:04   this and why they're not just using rate [TS]

00:51:05   so he describes two flaws with one of [TS]

00:51:08   the popular raid schemes known as raid 5 [TS]

00:51:10   raid 5 is the one where you have [TS]

00:51:11   multiple disks then one disk it's a [TS]

00:51:12   parity disk and if you lose one of the [TS]

00:51:14   disks you can rebuild its contents [TS]

00:51:16   so the key flaw with raid the points out [TS]

00:51:23   the first one is something called the [TS]

00:51:24   raid 5 right hole and the basic idea is [TS]

00:51:28   that whenever you update the data enter [TS]

00:51:30   what's called a raid stripe which is the [TS]

00:51:31   the data it's spread across all discs [TS]

00:51:33   you also have to update that parry disk [TS]

00:51:34   and that makes sense the whole point is [TS]

00:51:35   its you know this is derived from those [TS]

00:51:37   other pieces hero 2 disks you have to [TS]

00:51:39   update them both at the same time and [TS]

00:51:42   the problem is there's no way to update [TS]

00:51:45   more than one disk atomically if you [TS]

00:51:47   update the first disk update the second [TS]

00:51:49   is a powerful well when you power comes [TS]

00:51:52   back on you're halfway done and that's [TS]

00:51:55   bad and it's now [TS]

00:51:56   atomically means it's either all done or [TS]

00:51:58   not done at all all right [TS]

00:52:01   and the second thing he talks about is a [TS]

00:52:04   partial stripe right which is when [TS]

00:52:06   you're updating some data but since raid [TS]

00:52:09   cuts everything into equal sized slices [TS]

00:52:11   maybe you're not updating a full one of [TS]

00:52:13   those slices well at the end of this you [TS]

00:52:16   can update the parity bit that you know [TS]

00:52:18   that's derived from all those other [TS]

00:52:19   pieces so you have to read all the old [TS]

00:52:22   data even though if you updated just one [TS]

00:52:24   little piece of it and that's a big [TS]

00:52:26   performance hit because if you're just [TS]

00:52:27   updating one little piece why do I have [TS]

00:52:29   to read all these all its neighbors [TS]

00:52:30   basically to recompute the the parity [TS]

00:52:33   bit so I can write the parity thing back [TS]

00:52:35   so this performance problem and a [TS]

00:52:37   possible data graph problem now raid [TS]

00:52:39   vendors solve these things by with [TS]

00:52:41   hardware solutions so they put NVRAM in [TS]

00:52:43   their storage things but I said I even [TS]

00:52:44   if you pull the plug nvram is it's [TS]

00:52:46   non-volatile Ram when you pull the plug [TS]

00:52:48   on it's still there and it would sort of [TS]

00:52:49   journal to the NVRAM and say that's the [TS]

00:52:52   wrong term but it would keep track of [TS]

00:52:54   what it was doing so that when you put [TS]

00:52:55   the power back on it would know uh [TS]

00:52:56   actually I was in the middle of [TS]

00:52:58   something and the data that I was about [TS]

00:52:59   to write to these three or four disks [TS]

00:53:00   I've still got it here so let me finish [TS]

00:53:03   that right and that works because they [TS]

00:53:04   had some place where you could store the [TS]

00:53:06   stuff that wasn't a disk but also didn't [TS]

00:53:09   disappear when you pulled the plug now [TS]

00:53:13   as you can imagine I the one of the few [TS]

00:53:15   early in my career actually dealt with [TS]

00:53:17   raid a little bit and we dealt with [TS]

00:53:19   storage vendors that sold you a raid [TS]

00:53:20   solution with this that NVRAM type of [TS]

00:53:23   thing and the one problem I can remember [TS]

00:53:26   ever having with that raid box that we [TS]

00:53:29   dealt with I don't remember the vendor [TS]

00:53:30   but might have been NetApp I don't [TS]

00:53:31   remember [TS]

00:53:31   was it the NVRAM went bad and so this [TS]

00:53:35   this thing is supposed to save us from [TS]

00:53:37   all these problems and stop working and [TS]

00:53:41   believe me that makes your storage go [TS]

00:53:43   crazy like what's going on it says the [TS]

00:53:45   data is corrupted the disk checks out [TS]

00:53:46   and I did that the nvram was bad our [TS]

00:53:48   section of the end of yarra that's the [TS]

00:53:49   solution was the yank out the card and [TS]

00:53:51   put in a new car with a new Ram but but [TS]

00:53:52   that point a bunch of data had been [TS]

00:53:53   corrupted all right so that's not a [TS]

00:53:55   great solution so what the ZFS guys did [TS]

00:53:59   was try to address each one of these [TS]

00:54:00   problems they tried to come up with a [TS]

00:54:02   file system that had a way of atomically [TS]

00:54:05   updating things so either all their [TS]

00:54:08   updates happen or none of them happen [TS]

00:54:10   and they're not the first to do is [TS]

00:54:12   NetApp speaking of them or run the run [TS]

00:54:13   of the first vendors to popularize this [TS]

00:54:15   type of thing where the problem with [TS]

00:54:17   being in the middle of doing something [TS]

00:54:18   and then getting interrupted and not [TS]

00:54:21   knowing where the heck you are the [TS]

00:54:22   solution the best solution is not to say [TS]

00:54:24   okay we'll always know we're where we [TS]

00:54:25   are [TS]

00:54:26   we'll just keep track of it in this [TS]

00:54:27   thing that even if you pull the power [TS]

00:54:29   out we'll always be okay like the nvram [TS]

00:54:31   or something like that so we'll always [TS]

00:54:32   know where we were we'll never will [TS]

00:54:33   never lose track uh and but that doesn't [TS]

00:54:37   work because you know look what if the [TS]

00:54:38   nvram is bad or what you know it's [TS]

00:54:41   you're not protecting yourself you're [TS]

00:54:42   just you just moving the problem to [TS]

00:54:43   another place because then for example [TS]

00:54:45   when you're writing to the NVRAM now do [TS]

00:54:47   you need a system there to make sure all [TS]

00:54:49   the updates to NVRAM are atomic and it [TS]

00:54:51   just you know you're just chasing your [TS]

00:54:52   tail at that point so the system that [TS]

00:54:55   NetApp and several other vendors used [TS]

00:54:58   and also ZFS use is don't update data in [TS]

00:55:01   place that means if you have a place [TS]

00:55:03   where some data is and it's changing to [TS]

00:55:04   something different don't overwrite the [TS]

00:55:06   existing thing with the new thing [TS]

00:55:07   because if you die in the middle that [TS]

00:55:09   what you got is half the old thing and [TS]

00:55:11   half in the new thing and that's bad [TS]

00:55:12   right even if you know what you were [TS]

00:55:14   doing then you gotta like hold geez do I [TS]

00:55:15   have either the old one somewhere in its [TS]

00:55:17   entirety or the new one somewhere it's [TS]

00:55:19   an entire in its entirety so the old one [TS]

00:55:21   somewhere it's an entirety you probably [TS]

00:55:23   don't have in the new one well maybe [TS]

00:55:25   that's an NVRAM but you know again [TS]

00:55:27   you're just moving the problem or app so [TS]

00:55:29   the probably the solution is don't [TS]

00:55:30   update data in place when you make a [TS]

00:55:32   change to some data do it by writing the [TS]

00:55:34   new data in a totally different place [TS]

00:55:36   then the only thing you have to do [TS]

00:55:39   atomically is say when you're all done [TS]

00:55:40   writing everything to this totally new [TS]

00:55:42   place that had nothing in it that was [TS]

00:55:43   totally clean flip one little thing that [TS]

00:55:45   says stop looking at the old place and [TS]

00:55:47   start looking at the new place and [TS]

00:55:49   that's much easier to manage that tiny [TS]

00:55:52   little flip because you know if if [TS]

00:55:55   you're pointing to the old location [TS]

00:55:56   everything is fine if you're putting the [TS]

00:55:57   new location everything is fine and the [TS]

00:56:00   only way you could be pointing in [TS]

00:56:01   between is to say like I halfway updated [TS]

00:56:03   the location of where things were [TS]

00:56:05   pointing so that the second solution to [TS]

00:56:07   that is uh since the FS has Jack [TS]

00:56:08   checksums everywhere is that it will [TS]

00:56:11   know when it says okay we're looking for [TS]

00:56:13   this thing and I'll say it looks at it [TS]

00:56:14   and says well this this checksum doesn't [TS]

00:56:17   match because the power when [TS]

00:56:18   and the checksum is the old checksum and [TS]

00:56:21   the data is half the old date and half [TS]

00:56:22   the new data so where where do I go [TS]

00:56:26   whereas I have no idea what to do all I [TS]

00:56:27   know is that this point of this data is [TS]

00:56:29   bad because the checksum doesn't match [TS]

00:56:31   it but it doesn't know why the checksum [TS]

00:56:33   doesn't match and so one of the [TS]

00:56:35   solutions ZFS uses for that is to [TS]

00:56:37   redundantly store all of its metadata [TS]

00:56:40   this is actually option it's not all of [TS]

00:56:42   it but it can do something called ditto [TS]

00:56:44   blocks where it says metadata is so [TS]

00:56:46   important that thing we use to keep [TS]

00:56:47   track of where everything is that I want [TS]

00:56:49   to put this in multiple places on the [TS]

00:56:50   disk physically separated from each [TS]

00:56:52   other far away just to maximize our [TS]

00:56:54   chances of having a good copy of all [TS]

00:56:56   this data so it's really a [TS]

00:56:58   belt-and-suspenders approach don't [TS]

00:57:00   update data in place important data put [TS]

00:57:02   in multiple places widely separated [TS]

00:57:05   maybe on separate disks may be on [TS]

00:57:06   separate portions of the same disk ah so [TS]

00:57:09   that ideally there's the the purpose of [TS]

00:57:13   design is there should be nothing you [TS]

00:57:15   can do to a properly operating ZFS disk [TS]

00:57:18   in terms of you know yanking the power [TS]

00:57:21   or interrupting something the middle or [TS]

00:57:22   whatever the ever results in a disk that [TS]

00:57:25   is not in a consistent state it may not [TS]

00:57:27   be in the state you want it it may be [TS]

00:57:29   like well this right succeeded in this [TS]

00:57:31   one didn't but it will always be [TS]

00:57:32   consistent as far as the file system [TS]

00:57:34   concerned the file system will never go [TS]

00:57:35   somewhere and say I can't find keep [TS]

00:57:38   track of where all these blocks are for [TS]

00:57:39   this file there's supposed to be 20 [TS]

00:57:40   blocks and I only know where ten are or [TS]

00:57:42   I can't tell whether the new version of [TS]

00:57:45   the old version this file should be [TS]

00:57:46   shown fastest inconsistency may not seem [TS]

00:57:51   that interesting from an application [TS]

00:57:52   level because a perfectly consistent [TS]

00:57:54   file system can still be filled with [TS]

00:57:56   garbage data but as far as I'm concerned [TS]

00:57:57   that is the job of the file system to [TS]

00:57:59   make sure the file system is consistent [TS]

00:58:00   so one of the slogan t-shirts that the [TS]

00:58:05   ZFS guys had I should have found a link [TS]

00:58:07   to this to it and I think it said fsck [TS]

00:58:09   space you fsck is the UNIX utility for [TS]

00:58:13   sensor file system check for crawling [TS]

00:58:16   over your whole disk and looking at [TS]

00:58:18   everything and says this says should be [TS]

00:58:19   10 files are there 10 pods it shows to [TS]

00:58:21   be 20 blocks those file they're 20 [TS]

00:58:22   blocks it says the the the size of this [TS]

00:58:24   thing should be this size if I add up [TS]

00:58:25   all the blocks as equal that size and [TS]

00:58:27   all you know the date for this file [TS]

00:58:28   looks like it was from the 1800s that's [TS]

00:58:30   probably wrong [TS]

00:58:31   but checking the file system means [TS]

00:58:34   crawling all over it and trying to look [TS]

00:58:36   for inconsistencies and the idea but ZFS [TS]

00:58:38   is there is no fsck the the file system [TS]

00:58:41   is always consistent on disk always uh [TS]

00:58:45   and there's there should be no possible [TS]

00:58:47   way that it can't be because you see [TS]

00:58:48   everything is either in the old state or [TS]

00:58:49   in the new state and we can't actually [TS]

00:58:50   update things atomically because we [TS]

00:58:52   don't up they did in place and because [TS]

00:58:53   we do also redundancy stuff and on top [TS]

00:58:55   of that we put check sums with [TS]

00:58:57   everything so that if the hardware goes [TS]

00:58:59   bad that's we're going to take that onto [TS]

00:59:00   and say the file system seems consistent [TS]

00:59:03   but some data is wrong and ZFS will will [TS]

00:59:07   will actually take it upon itself to say [TS]

00:59:09   look this data is on three is in three [TS]

00:59:12   different locations on three different [TS]

00:59:13   disks but I'm one of the disks that [TS]

00:59:14   checksum is bad another to this the [TS]

00:59:16   checksum is correct so it will take the [TS]

00:59:19   good data off one of the good discs and [TS]

00:59:20   put it back onto the bad disk so it can [TS]

00:59:23   sort of heal itself you know due to a [TS]

00:59:26   hardware bit flip type of errors and it [TS]

00:59:27   will try to write to a new location so [TS]

00:59:29   it doesn't write over the same possibly [TS]

00:59:31   bad bits it's a file system that's [TS]

00:59:34   actually taking responsibility for the [TS]

00:59:35   data that it's storing and it's saying [TS]

00:59:37   I'm our job is to make sure I you know [TS]

00:59:38   that my bookkeeping is never all messed [TS]

00:59:41   up and I have no idea what's going on so [TS]

00:59:42   basically none of those little red error [TS]

00:59:44   message isn't Disk Utility no no F SDK [TS]

00:59:47   to crawl of your whole disk sleep is [TS]

00:59:48   okay it's always okay and if you have a [TS]

00:59:50   hardware failure I'm going to tell you [TS]

00:59:52   about it and if I can fix it because I [TS]

00:59:53   have redundant copies of the data I will [TS]

00:59:55   do so so that's sort of what's exciting [TS]

00:59:59   about ZFS [TS]

00:59:59   about ZFS [TS]

01:00:00   and nacho I didn't go into ZFS is [TS]

01:00:01   solution for a raid and going across [TS]

01:00:04   multiple disks but yes they have a thing [TS]

01:00:05   where you can put multiple but that's a [TS]

01:00:07   behind-the-scenes thing and it doesn't [TS]

01:00:08   affect the way that ZFS works from the [TS]

01:00:10   user or the developer standpoint yeah [TS]

01:00:13   that's it esoteric like debating you [TS]

01:00:16   know it is easier to resize volumes and [TS]

01:00:17   raid versus some of the other solutions [TS]

01:00:19   and NetApp net up has a similar type of [TS]

01:00:21   file system called waffle I don't know [TS]

01:00:23   what they call now with those the [TS]

01:00:24   original name whose a write anywhere [TS]

01:00:26   file layout or log structured file [TS]

01:00:29   systems those are the ones but it's a we [TS]

01:00:30   don't update in place we always put data [TS]

01:00:32   in new locations and there are all sorts [TS]

01:00:33   of performance consequences of this but [TS]

01:00:37   basically that's the the pitch for for [TS]

01:00:40   ZFS and for a lot of modern file systems [TS]

01:00:42   a btrfs I don't even know that how [TS]

01:00:45   that's if that's how you pronounce that [TS]

01:00:46   but btrfs is a similar type of file [TS]

01:00:48   system in the open source space [TS]

01:00:50   like I said net app has its own file [TS]

01:00:52   system sets I think it's still suing son [TS]

01:00:55   and now Oracle over the patents involved [TS]

01:00:57   in that but way back in the corner is [TS]

01:00:59   poor old HFS+ that has none of this but [TS]

01:01:01   updates all its data in place that has [TS]

01:01:03   all these sorts of weird design [TS]

01:01:04   decisions that made sense when we were [TS]

01:01:05   on floppy disks and 10 megabyte hard [TS]

01:01:07   drives wouldn't make no sense now and [TS]

01:01:09   it's had many things add to it try to [TS]

01:01:11   allow to support the features of a [TS]

01:01:13   modern unix that you know hard links [TS]

01:01:16   symlinks unicode wasn't grafted on but [TS]

01:01:21   the implementation of unicode has come [TS]

01:01:23   under scrutiny from from linus Toral's [TS]

01:01:26   in particular if you want to read about [TS]

01:01:27   how he hates hfs+ because of the way it [TS]

01:01:29   does unicode normalization or what new [TS]

01:01:31   unicode normalization is you can google [TS]

01:01:32   that but it doesn't have any of these [TS]

01:01:36   things about protecting your data or our [TS]

01:01:38   logical volume management or not [TS]

01:01:42   corrupting itself all that stuff so the [TS]

01:01:44   world is move on but H of s plus is not [TS]

01:01:46   so for years and years I've been saying [TS]

01:01:47   look Apple you gotta do something about [TS]

01:01:48   the file system every other file system [TS]

01:01:51   that's out there is better in one or [TS]

01:01:55   more important regards in hfs+ don't [TS]

01:01:57   keep adding things to HFS+ it's not the [TS]

01:01:58   way to do it you're sacrificing [TS]

01:02:00   performance and features and and don't [TS]

01:02:02   you feel like you have to take [TS]

01:02:03   responsibility for the data that's on [TS]

01:02:04   the disk I didn't really care what it [TS]

01:02:05   was they write their own file system if [TS]

01:02:07   they adopt [TS]

01:02:08   ZFS they've dot btrfs they anything you [TS]

01:02:11   know just [TS]

01:02:13   don't keep it around forever and ZFS the [TS]

01:02:18   the move does ZFS look like it was going [TS]

01:02:19   to be that but that kind of fell through [TS]

01:02:20   so it's like alright well what's what's [TS]

01:02:21   your new plan and so far no new plan but [TS]

01:02:24   in Lion they introduced something called [TS]

01:02:26   core storage which is apples crack at [TS]

01:02:28   logical volume management and they did [TS]

01:02:29   that because they had to do it to [TS]

01:02:30   support the new whole disk encryption [TS]

01:02:32   thing and I speculated in the line [TS]

01:02:34   review that perhaps core storage could [TS]

01:02:37   be laying the foundation for a new file [TS]

01:02:41   system from Apple that would be they [TS]

01:02:44   would use some of that work and that [TS]

01:02:47   would go some way towards addressing [TS]

01:02:48   some of the concerns with with hfs+ but [TS]

01:02:51   that's all just speculation and I don't [TS]

01:02:53   know if that's going to happen and I [TS]

01:02:55   still have not checked I'm assuming [TS]

01:02:57   mountain lion continues hfs+ that I [TS]

01:02:58   literally have not checked so I can not [TS]

01:03:00   violate my NDA and saying I don't know [TS]

01:03:02   if it does but I will be very surprised [TS]

01:03:04   if it doesn't because that seems like [TS]

01:03:05   something I would have heard about in [TS]

01:03:06   blogs many times over now all this wraps [TS]

01:03:10   all the way around believe it or not to [TS]

01:03:12   Microsoft didn't see that one coming [TS]

01:03:15   didn't know I would I'm I'm just shocked [TS]

01:03:18   that you would go through this is the [TS]

01:03:20   reason this is even in my show notes we [TS]

01:03:22   got interrupted by mountain lion and the [TS]

01:03:24   show I missed in all those stuff because [TS]

01:03:26   Microsoft made an announcement of [TS]

01:03:28   something they called they called re FS [TS]

01:03:30   capital R lowercase e capital FS high [TS]

01:03:34   which stands for resilient file system [TS]

01:03:36   and so Microsoft has a similarly [TS]

01:03:39   checkered history of file system stuff [TS]

01:03:41   they were supposed to come out with win [TS]

01:03:42   FS which really wasn't a file system but [TS]

01:03:44   was supposed to have a lot of features [TS]

01:03:46   of logical volume management and [TS]

01:03:48   searching for data and metadata and all [TS]

01:03:50   sorts of stuff but like there was [TS]

01:03:51   confusion about that really it's not a [TS]

01:03:53   file system it's just a layer on top of [TS]

01:03:55   our existing file system but then they [TS]

01:03:56   got canceled that's part of the Longhorn [TS]

01:03:59   just a mess [TS]

01:04:01   so many years of Microsoft has been [TS]

01:04:03   using NTFS which is the file system [TS]

01:04:05   developer windows empty which is not a [TS]

01:04:07   terrible file system it certainly [TS]

01:04:10   started out it's like much more advanced [TS]

01:04:12   than Apple's file systems and it's [TS]

01:04:14   mostly still more advanced than HFS+ has [TS]

01:04:18   better performance characteristics [TS]

01:04:19   there's some weird things about it it's [TS]

01:04:21   certainly not of the vintage of ZFS or [TS]

01:04:24   any of the other more modern [TS]

01:04:27   systems but it's worlds better than fat [TS]

01:04:29   or fat32 but any right that's what [TS]

01:04:33   Microsoft is is using these days and [TS]

01:04:36   Microsoft true to form is usually pretty [TS]

01:04:39   good about seeing where it has a [TS]

01:04:41   technical failing and addressing it so [TS]

01:04:43   for example they saw when Java was [TS]

01:04:46   introduced and memory manage languages [TS]

01:04:49   were coming out that having a platform [TS]

01:04:51   that's entirely based on C and C++ is [TS]

01:04:53   kind of it might be seen as not keeping [TS]

01:04:58   up with the time so in response that [TS]

01:05:00   they made the dotnet platform in a [TS]

01:05:02   common language runtime and it's C sharp [TS]

01:05:04   language and you know they saw a place [TS]

01:05:06   where they were falling behind [TS]

01:05:07   technically and they addressed it in a [TS]

01:05:08   big long multi-year plan that took a [TS]

01:05:10   long time to execute and arguably they [TS]

01:05:13   did a bad job and certain aspects of it [TS]

01:05:14   but the point is they're they're [TS]

01:05:15   reacting to trends and trying to keep up [TS]

01:05:19   so here they are looking at the Frost's [TS]

01:05:20   and saying NTFS it's not bad but the [TS]

01:05:23   world the file systems has moved on we [TS]

01:05:24   need something better and so riaf asses [TS]

01:05:27   their thing it's going to be introduced [TS]

01:05:28   as part of Windows Server 8 according to [TS]

01:05:30   Microsoft and this is the way Microsoft [TS]

01:05:32   does all its introductions it tends to [TS]

01:05:34   release the file system on the server [TS]

01:05:35   platform first because like those the [TS]

01:05:37   guys who are interested in file systems [TS]

01:05:38   and they can kind of test it out there [TS]

01:05:40   before it comes down to the consumers [TS]

01:05:41   like I don't remember when NTFS you [TS]

01:05:43   might know this as that X Windows nerd [TS]

01:05:45   but an NTFS started in NT and Windows [TS]

01:05:49   was stuck with fat for the longest time [TS]

01:05:50   and it was at Windows 2000 when NTFS [TS]

01:05:52   became the default yeah I think you're [TS]

01:05:54   right about that that sounds right to me [TS]

01:05:56   as far as being the default or you do [TS]

01:05:58   mean something that you could use I know [TS]

01:06:01   that I know for sure that it was in [TS]

01:06:03   Windows 2000 and of course our XP and [TS]

01:06:05   everything after that and that it [TS]

01:06:07   superseded fat but I'm trying to [TS]

01:06:10   remember when that actually happened um [TS]

01:06:14   an action started calling it like NTFS [TS]

01:06:18   version 5 back in the 2000 time period [TS]

01:06:22   they say good job though because like if [TS]

01:06:24   you bought Windows like you ideally you [TS]

01:06:27   wouldn't even know that like the next PC [TS]

01:06:29   you bought had NTFS on yeah right like [TS]

01:06:31   it wouldn't be an issue and like I mean [TS]

01:06:34   you reinstall windows I don't know if [TS]

01:06:35   they had a cross converter that said it [TS]

01:06:37   looks like you have fat32 bit so you [TS]

01:06:38   want us to converge on it convert you [TS]

01:06:40   definitely could [TS]

01:06:41   I'm Apple tried to do that too by the [TS]

01:06:43   way I don't know you weren't perhaps [TS]

01:06:45   well maybe do you remember that when [TS]

01:06:46   hfs+ suki datta sure you could convert [TS]

01:06:49   your age of s plus your age of s disc [TS]

01:06:51   date reverse plus I mean I started using [TS]

01:06:53   Mac's back in the mid-80s so it there [TS]

01:06:56   was never time when I didn't have one [TS]

01:06:57   but yeah I remember when you could [TS]

01:06:58   convert it and it was a very say was a [TS]

01:07:00   very serious decision that you had to [TS]

01:07:02   make it was it was the real deal [TS]

01:07:04   and since HFS like HFS+ was prone to [TS]

01:07:07   corrupt its metadata structures that [TS]

01:07:09   conversion process could totally destroy [TS]

01:07:10   yeah it was a sinner had to have really [TS]

01:07:12   good backups yeah I mean is like I'll [TS]

01:07:15   just I'll just update my volume format [TS]

01:07:17   in place and what could go wrong all [TS]

01:07:19   right here's here's what it says [TS]

01:07:21   Microsoft released version 3 of NTFS [TS]

01:07:23   sometimes incorrectly called NTFS 5 in [TS]

01:07:26   relation to the kernel version number as [TS]

01:07:27   part of Windows 2000 this introduced [TS]

01:07:30   disk quotas provided by quota advisor [TS]

01:07:31   file system level encryption sparse [TS]

01:07:34   files and reparse points sparse files [TS]

01:07:37   allow for the efficient storage of data [TS]

01:07:39   sets etc yeah this is this is when it [TS]

01:07:44   was introduced into 2000 and I think [TS]

01:07:46   weren't there also some may be already [TS]

01:07:49   talked about this a little bit I'm [TS]

01:07:50   trying to remember what the file system [TS]

01:07:53   size limits were remember I used to have [TS]

01:07:55   those and with I owed by I don't know [TS]

01:07:59   what the limits are Windows BIA there [TS]

01:08:00   was some very low limits on on disk size [TS]

01:08:03   that disk were quickly outpacing how [TS]

01:08:05   large a volume could be in fat and in [TS]

01:08:08   fat32 extended that but you were still [TS]

01:08:09   bumping up against the limits and DFS [TS]

01:08:11   even NTFS might have had its limits [TS]

01:08:13   bumped a few times that so something [TS]

01:08:14   that all file systems tend to have [TS]

01:08:15   because when you design a file system [TS]

01:08:16   you're like okay and then this is where [TS]

01:08:18   I'm going to store the maximum number of [TS]

01:08:21   whatever's and then whatever is can each [TS]

01:08:23   point to a certain number of whatever is [TS]

01:08:25   knows whatever you know you eventually [TS]

01:08:26   add it up and you say ok given that I [TS]

01:08:28   use 16 bits for this value or 32 bits [TS]

01:08:31   for this value on disk and I multiply [TS]

01:08:32   that out what is the maximum maximum [TS]

01:08:35   file size maximum volume size you make [TS]

01:08:38   decisions when you're storing stuff on [TS]

01:08:39   disk and those decisions have [TS]

01:08:40   repercussions in the capacity of your [TS]

01:08:44   file system and as you can imagine [TS]

01:08:46   people who made those decisions in 1981 [TS]

01:08:49   or whatever we're like oh let's just use [TS]

01:08:50   16 bits for this because every every [TS]

01:08:53   every every byte counted [TS]

01:08:54   and maybe add 16-bit registers on your [TS]

01:08:56   CPU and it just worked out well you know [TS]

01:08:58   even 32 bits like or insert 32 bits [TS]

01:09:00   around 486 is now let's just use 32 bits [TS]

01:09:02   it works nice and if we use something [TS]

01:09:04   longer then you got a chop it up and [TS]

01:09:06   it's just a pain in my butt and I'm not [TS]

01:09:07   going to do that and by the way I've all [TS]

01:09:09   put I found a neat little article on [TS]

01:09:11   this I'll put in the show notes we want [TS]

01:09:12   to mention how to get to the show notes [TS]

01:09:13   you go to five by five - TV slash [TS]

01:09:15   hypercritical slash 56 and you'll see [TS]

01:09:18   all of the links that we have we've [TS]

01:09:21   mentioned and all the things that John [TS]

01:09:23   stated painted painstakingly adds to the [TS]

01:09:26   show they're all there you can subscribe [TS]

01:09:28   to those in our RSS reader by the way [TS]

01:09:31   and you can follow along at home that [TS]

01:09:33   way and thanks very much to help [TS]

01:09:35   spot.com the best helpdesk software in [TS]

01:09:38   the business they're sponsoring show [TS]

01:09:40   notes so that but there's a neat little [TS]

01:09:43   article that did you ever were you ever [TS]

01:09:44   a user of Windows 2000 or NT to find a [TS]

01:09:48   user I had one on your desk that ya D [TS]

01:09:52   work on whether they were happy about it [TS]

01:09:55   or not frequently had a a Windows VM of [TS]

01:10:00   some kind to do browser testing and ie [TS]

01:10:04   on Windows but I've never actually used [TS]

01:10:06   one as my main machine but yes I've [TS]

01:10:08   interacted with Windows 2000 and T there [TS]

01:10:12   is a great I don't know if you ever [TS]

01:10:13   remember this feature it's not so much a [TS]

01:10:15   filesystem feature but it was something [TS]

01:10:17   that was supported that you could you [TS]

01:10:18   could compress individual folders not [TS]

01:10:22   zip them but you could you could [TS]

01:10:24   compress them so that anything that you [TS]

01:10:26   put in that folder would be [TS]

01:10:27   automatically compressed in that [TS]

01:10:30   directory so that you could use so you [TS]

01:10:32   said well just just compress this [TS]

01:10:34   directory and none of these other ones [TS]

01:10:35   and it didn't seem to impact performance [TS]

01:10:37   too much yeah that's something that [TS]

01:10:39   actually Apple added a trust plus [TS]

01:10:41   speaking up and something of the ZFS [TS]

01:10:42   supports as well as a you know [TS]

01:10:45   encryption and compression of using [TS]

01:10:47   several different algorithms that's [TS]

01:10:48   that's another modern filesystem feature [TS]

01:10:50   that for example fat didn't have an H [TS]

01:10:52   best bliss didn't have it for a long [TS]

01:10:54   time and they just recently added can [TS]

01:10:55   you currently do that on Mac OS 10 can [TS]

01:10:58   you compress just one folder I don't [TS]

01:11:01   know if the mechanism form triggering [TS]

01:11:03   the compression is accessible in a nice [TS]

01:11:05   way but to compress a full [TS]

01:11:07   what you would basically do is compress [TS]

01:11:08   all the files in it all the files of [TS]

01:11:10   those folders contain you know on the [TS]

01:11:12   way down so there's not you know the [TS]

01:11:13   idea of compressing a folder is [TS]

01:11:14   basically just compressed the folders [TS]

01:11:16   contents and tons of stuff is compressed [TS]

01:11:18   on your disk right now if you're using a [TS]

01:11:20   lion or Snow Leopard which I think is [TS]

01:11:24   where this feature is introduced hfs+ [TS]

01:11:25   does support transparent encryption it's [TS]

01:11:27   just that it's not there's no [TS]

01:11:28   right-click where you can like in [TS]

01:11:30   Windows but you could say compress it [TS]

01:11:31   it's they tend to compress stuff that's [TS]

01:11:34   only going to be read like the [TS]

01:11:35   applications or parts of the operating [TS]

01:11:37   system and I think in if you look at my [TS]

01:11:39   my I'm getting about say snow line Snow [TS]

01:11:43   Leopard review you can find some links [TS]

01:11:47   to some tools that will let you see [TS]

01:11:48   which files are compressed or not but [TS]

01:11:51   what could that information does you I'm [TS]

01:11:52   not sure because it's like well can I [TS]

01:11:54   decompress it not easily there's no real [TS]

01:11:57   UI for that can I compress this file not [TS]

01:11:59   really you know it's clearly a feature [TS]

01:12:01   that Apple's using to optimize its its [TS]

01:12:03   operating system but it's not user [TS]

01:12:05   accessible so getting back to Rio fest [TS]

01:12:08   these are all things we talked about [TS]

01:12:10   that NTFS you know had that fat didn't [TS]

01:12:12   have but even NTFS Microsoft that like [TS]

01:12:14   was behind because it doesn't do all [TS]

01:12:16   those fancy things that I talked about [TS]

01:12:17   ZFS doing so I link to Microsoft's [TS]

01:12:20   explanation of r EF s on MSDN page which [TS]

01:12:23   is very good and I just highlighted a [TS]

01:12:26   few sections of it here the key goals of [TS]

01:12:29   re FS so the first goal is they wanted [TS]

01:12:34   to look like NTFS to the software which [TS]

01:12:36   is typical Microsoft fashion they like [TS]

01:12:38   backward compatible I so your software [TS]

01:12:39   should have no idea that its operating [TS]

01:12:41   on re f s they're trying to maintain all [TS]

01:12:44   the features that NTFS supports all the [TS]

01:12:46   API is the NTFS supports like all the [TS]

01:12:48   filesystem API is this probably this I [TS]

01:12:50   read this a few times like what are they [TS]

01:12:51   even talking about but I realized if [TS]

01:12:53   you're not in the UNIX mindset it kind [TS]

01:12:54   of makes sense and in UNIX minds that [TS]

01:12:55   you're like why would a new file system [TS]

01:12:57   change the API is that he used like I [TS]

01:12:59   use open you know and RF open or [TS]

01:13:02   whatever and read it's like that we're [TS]

01:13:04   kind of used to in UNIX there's one set [TS]

01:13:07   of file i/o operations and an API is [TS]

01:13:09   that works with all file systems and [TS]

01:13:11   it's up to the file system driver to [TS]

01:13:12   abstract that stuff out but apparently [TS]

01:13:14   in the world of Windows and I'm sure in [TS]

01:13:16   the world of Mac do this certain api's [TS]

01:13:17   that are specific to certain file system [TS]

01:13:19   so Microsoft is being [TS]

01:13:21   explicit here and saying look the exact [TS]

01:13:22   same API is that you used interact with [TS]

01:13:24   your NT bestest now those exact api's [TS]

01:13:26   will work with r EF s so that's very [TS]

01:13:27   important to them they don't want to [TS]

01:13:29   break backward compatibility they wanted [TS]

01:13:31   to look just like an NTFS disk but [TS]

01:13:33   that's boring so the first key goal [TS]

01:13:35   that's not boring is verify and [TS]

01:13:37   autocorrect data that should sound very [TS]

01:13:38   familiar all right they want to be able [TS]

01:13:41   to tell that if the data is correct they [TS]

01:13:45   write it and then six months later you [TS]

01:13:47   should look at it and say these the [TS]

01:13:49   checksum doesn't match data is bad and [TS]

01:13:51   autocorrect means the type of thing that [TS]

01:13:53   we're talking about ZFS where it's like [TS]

01:13:55   alright this data is bad do we have [TS]

01:13:56   another copy this data somewhere and the [TS]

01:13:57   checksum does match on that one well [TS]

01:13:59   then let's put that data back over here [TS]

01:14:00   now they're both good again one of the [TS]

01:14:04   things I bowled out here is metadata [TS]

01:14:05   must not be written in place also sounds [TS]

01:14:07   familiar don't do you know what they [TS]

01:14:09   call torn writes where you're halfway [TS]

01:14:11   through over writing something you just [TS]

01:14:12   end up with garbage is an interesting [TS]

01:14:17   one which is kind of a dig at ZFS don't [TS]

01:14:19   assume that disk checking algorithm is [TS]

01:14:20   in fact this is not a dig a little digs [TS]

01:14:22   come later don't assume that just [TS]

01:14:23   checking algorithms in particular can [TS]

01:14:25   scale to the size of an entire file [TS]

01:14:26   system I think that's saying you're not [TS]

01:14:30   gonna be able to fsck a four terabyte [TS]

01:14:32   disk in a couple you like it you know [TS]

01:14:34   how long it takes to check every single [TS]

01:14:35   button on a disk to make sure that the [TS]

01:14:37   structures are consistent it's kind of [TS]

01:14:38   ridiculous [TS]

01:14:39   never take the filesystem offline this [TS]

01:14:42   is a very interesting one that on the [TS]

01:14:44   game ZFS says is they want to keep the [TS]

01:14:47   entire file system online when things [TS]

01:14:49   are going wrong without having to take [TS]

01:14:50   anything off line and provide end-to-end [TS]

01:14:53   resiliency bla bla bla bla so it reads [TS]

01:14:55   almost like a feature list from ZFS like [TS]

01:14:58   all that stuff the ZFS has talked about [TS]

01:14:59   we're going to do our version of it and [TS]

01:15:01   presumably we'll do it better or in a [TS]

01:15:02   more interesting manner or we will learn [TS]

01:15:04   the lessons of ZFS and then so they have [TS]

01:15:10   a separate section for key features [TS]

01:15:11   because those are the goals I kind of [TS]

01:15:12   started talking about the features but [TS]

01:15:13   their their features are they list here [TS]

01:15:16   are metadata integrity with checksums so [TS]

01:15:19   they're just talking about the thing [TS]

01:15:23   that keeps track of where your stuff is [TS]

01:15:24   and I'm talking about the data itself [TS]

01:15:25   then they say integrity streams [TS]

01:15:27   providing optional user data integrity [TS]

01:15:29   so they've separated those things out [TS]

01:15:30   same metadata integrity will make sure [TS]

01:15:32   that everything is consistent on disk so [TS]

01:15:34   we [TS]

01:15:34   to protect our metadata our metadata is [TS]

01:15:35   our bookkeeping of where stuff is and [TS]

01:15:37   that we will checks on the hell out of [TS]

01:15:39   so that we can tell if that's messed up [TS]

01:15:40   because if that's messed up all bets are [TS]

01:15:41   off we could be reading in the wrong [TS]

01:15:42   place we could lose track of files or [TS]

01:15:44   check something that all the time but [TS]

01:15:46   they're hedging their bets on data [TS]

01:15:48   integrity they're going to say okay well [TS]

01:15:49   that's optional the actual contents of [TS]

01:15:52   your file we can check some notes for [TS]

01:15:53   you or we can not and why would you not [TS]

01:15:55   want to this is one of the complaints [TS]

01:15:57   against EFS is that if you check some [TS]

01:16:00   every single piece of data that means [TS]

01:16:03   you have to make a computation for every [TS]

01:16:05   little piece of data you put on disk and [TS]

01:16:06   every time you read it from disk you [TS]

01:16:07   have to redo that computation and [TS]

01:16:08   compare with a checksum which you also [TS]

01:16:09   have to read from the disk and make sure [TS]

01:16:11   they're correct and so it's it's [TS]

01:16:13   potentially CPU intensive using massive [TS]

01:16:15   amounts of Io ZFS struggled to convince [TS]

01:16:19   everyone that CPUs are so fast now with [TS]

01:16:22   multi-core it's not a problem and if it [TS]

01:16:26   you know it does take some cpu capacity [TS]

01:16:28   but what would you rather use your CPU [TS]

01:16:30   stuff for it I'm surprised they had such [TS]

01:16:31   a hard time arguing this because to my [TS]

01:16:33   mind it's kind of like the speed versus [TS]

01:16:36   correctness thing like what wouldn't be [TS]

01:16:37   much faster we didn't do those check [TS]

01:16:38   sums like yeah but then maybe your data [TS]

01:16:40   isn't correct so what do you want [TS]

01:16:41   correctness or speed and in programming [TS]

01:16:43   speed without correctness is pointless I [TS]

01:16:46   can make you an infinitely fast program [TS]

01:16:47   that's incorrect because it just [TS]

01:16:48   immediately exits all right [TS]

01:16:50   correctness has got to be number one so [TS]

01:16:52   the ZFS mindset is correctness above all [TS]

01:16:55   check sums on everything that's the way [TS]

01:16:57   it has to be Microsoft is still hedging [TS]

01:16:59   its bets saying we know that some of you [TS]

01:17:00   may not be ready to sacrifice your CPU [TS]

01:17:03   cores for the sake of data integrity and [TS]

01:17:06   we will leave that up to you if you [TS]

01:17:08   don't care if you're you know maybe it's [TS]

01:17:09   just scratch data and you don't care if [TS]

01:17:11   it's corrupt and if it's corrupt you'll [TS]

01:17:12   just wipe the whole disk because it's [TS]

01:17:13   just a cache in a CDN or something so [TS]

01:17:14   they're trying to be flexible here [TS]

01:17:16   metadata integrity always data integrity [TS]

01:17:19   sometimes allocate on write [TS]

01:17:22   transactional model for robust disk [TS]

01:17:23   updates that's jargon for the [TS]

01:17:26   transactional disk updates where you [TS]

01:17:28   don't overwrite existing data right in a [TS]

01:17:30   new place and either it's all written or [TS]

01:17:31   it's not written storage pooling and [TS]

01:17:33   virtualization Microsoft has another [TS]

01:17:35   thing called storage spaces that they [TS]

01:17:37   introduced a little bit before re FS was [TS]

01:17:40   announced it's their layer 4 [TS]

01:17:42   virtualizing storage I'm multiple discs [TS]

01:17:44   look like a single logical volume [TS]

01:17:47   disc scrubbing for protecting protecting [TS]

01:17:49   against Leyton disc errors the scrubbing [TS]

01:17:51   is the type of thing we're like every [TS]

01:17:52   once in a while why don't you check all [TS]

01:17:54   those check sums that I told you to put [TS]

01:17:55   on the data at because if you know if [TS]

01:17:57   you don't have occasion to ever read [TS]

01:17:59   that data you won't know it's bad so [TS]

01:18:00   maybe just check it see if it's good and [TS]

01:18:02   if it's bad get a good copy from [TS]

01:18:03   somewhere so this looks pretty good [TS]

01:18:06   compared to the feature set of ZFS it's [TS]

01:18:10   kind of like how I wanna in my mind I [TS]

01:18:11   see that like how does c-sharp stack up [TS]

01:18:13   against Java and how does we have a [TS]

01:18:15   stack up against the FS Java and ZFS [TS]

01:18:17   were the first ones also both [TS]

01:18:18   incidentally from Sun interestingly and [TS]

01:18:20   Microsoft gets the benefit of waiting [TS]

01:18:24   and seeing how that shakes out so ZFS [TS]

01:18:25   and Java like do their thing and then [TS]

01:18:28   Microsoft says well what do people not [TS]

01:18:29   like about Java what people don't like [TS]

01:18:31   about ZFS how can we do it better we're [TS]

01:18:33   going to do our own thing and they're [TS]

01:18:34   making some big advances there but how [TS]

01:18:35   can we do it better and so they've made [TS]

01:18:38   some slightly different decisions and [TS]

01:18:39   their implementation decisions are [TS]

01:18:40   different of course making compatible [TS]

01:18:42   than TFS and using old API so my [TS]

01:18:44   understanding is they're actually using [TS]

01:18:45   the top layer API literally as is with [TS]

01:18:49   all of its timings and they made a big [TS]

01:18:51   point of like you know for [TS]

01:18:53   multi-threaded programming stuff in [TS]

01:18:54   terms of the timings of you know the [TS]

01:18:57   code paths through the storage stack [TS]

01:18:59   they want that to be exactly the same as [TS]

01:19:01   in the TFS and only they'll chop off [TS]

01:19:02   only the backend when you're right into [TS]

01:19:04   the actual storage so everything from [TS]

01:19:05   the application perspective should be [TS]

01:19:06   bug for bug compatible with with the [TS]

01:19:09   NTFS storage layer compared to ZFS of [TS]

01:19:12   trying to think of how is there anything [TS]

01:19:14   there missing ditto blocks they kind of [TS]

01:19:18   have that in storage spaces where you [TS]

01:19:20   can put data in multiple disks but [TS]

01:19:21   dental blocks is in a block level where [TS]

01:19:22   you can just say make three copies of [TS]

01:19:24   this kind of disk because it's really [TS]

01:19:26   important this type of metadata or [TS]

01:19:27   whatever ZFS is really good about [TS]

01:19:29   putting important metadata in multiple [TS]

01:19:31   location and then protecting itself and [TS]

01:19:32   there's nothing really equivalent to ZFS [TS]

01:19:35   is solution to raid which it calls raid [TS]

01:19:38   Z and as raid zine raid Z - it's a [TS]

01:19:40   different way of arranging stuff [TS]

01:19:42   multiple disks but again storage spaces [TS]

01:19:43   but Microsoft is is going more to the [TS]

01:19:45   model where they're logical volume [TS]

01:19:46   manager is a separate layer from their [TS]

01:19:49   file system whereas ZFS combines them [TS]

01:19:52   into one because I think their [TS]

01:19:53   advantages in doing so I think there are [TS]

01:19:55   advantages of doing so but people don't [TS]

01:19:57   like that but layering and I think [TS]

01:19:59   Microsoft people Microsoft always he [TS]

01:20:01   didn't like [TS]

01:20:01   luring too much either see what else [TS]

01:20:05   we've got our us while we do our second [TS]

01:20:07   sponsor while you do your last-minute [TS]

01:20:10   research okay [TS]

01:20:11   it's MailChimp calm longtime sponsor and [TS]

01:20:14   I love these guys he's the heck out of [TS]

01:20:15   these guys easy mail newsletters that's [TS]

01:20:17   the easy way to think of it you want to [TS]

01:20:19   do a newsletter you do MailChimp com [TS]

01:20:21   that's it you go there you sign up you [TS]

01:20:24   send 12,000 emails to 2,000 people you [TS]

01:20:27   do that every month and you pay nothing [TS]

01:20:29   sounds pretty cool right you want [TS]

01:20:32   integrated into your interior iOS app [TS]

01:20:35   they give you all the code you need to [TS]

01:20:37   just do that same thing for Android apps [TS]

01:20:39   they also have their own mobile apps so [TS]

01:20:42   that if you're on the go [TS]

01:20:43   you want to see our people reading the [TS]

01:20:45   newsletters people following along with [TS]

01:20:47   the campaign they've got really really [TS]

01:20:49   awesome [TS]

01:20:50   it's called MailChimp mobile again all [TS]

01:20:52   this is free they've got integration [TS]

01:20:55   with stuff like woo foof you use them [TS]

01:20:56   for forums they've got integration with [TS]

01:20:58   Survey Monkey is if you want people to [TS]

01:20:59   do surveys they just put up some really [TS]

01:21:02   cool new videos just in the last couple [TS]

01:21:05   days showing how MailChimp works how [TS]

01:21:07   easy it is to use they've got guides for [TS]

01:21:09   everything breaking down every single [TS]

01:21:11   topic everything you want to do and make [TS]

01:21:14   with MailChimp they just make it easy to [TS]

01:21:16   do that so it's really really cool [TS]

01:21:20   they've got tons of resources and yeah [TS]

01:21:22   if you if you ramp things up you you [TS]

01:21:26   have more than 2000 subscribers you get [TS]

01:21:28   a pay but it's not expensive [TS]

01:21:30   they keep it really really cheap I mean [TS]

01:21:33   you know you're talking about like like [TS]

01:21:34   10 bucks a month and up so check it out [TS]

01:21:39   MailChimp com great guys great [TS]

01:21:41   supporters of 5x5 and if you need to do [TS]

01:21:44   anything with newsletters this is this [TS]

01:21:46   is the place to go alright with that I [TS]

01:21:54   was making sure wasn't muted okay [TS]

01:21:56   someone in the chat room posted a link [TS]

01:21:59   to entry from bond Wix blog called our [TS]

01:22:03   ramp and layering violation hmm Andrew [TS]

01:22:06   Morton someone in the Linux community a [TS]

01:22:09   bigwig limo security made a complaint to [TS]

01:22:11   when ZFS was first being paraded around [TS]

01:22:15   the [TS]

01:22:15   seen that it was a violet a layering [TS]

01:22:18   violation that you're supposed to have [TS]

01:22:19   the file system which just manages you [TS]

01:22:21   know a single volume and then you have [TS]

01:22:22   logical volume management which manages [TS]

01:22:23   multiple volumes that and putting it [TS]

01:22:25   together into one big ball of mud was [TS]

01:22:26   bad you're supposed to have separate [TS]

01:22:28   layers then you can change the layers [TS]

01:22:29   independently to get the best logical [TS]

01:22:30   while you manager with the best file [TS]

01:22:32   system and not surprisingly Jeff bond [TS]

01:22:35   works response was that we think that [TS]

01:22:38   combining them has advantages and they [TS]

01:22:41   go he goes into that in the post I [TS]

01:22:42   backed out why had the shownotes link [TS]

01:22:43   now and I actually did a blog post about [TS]

01:22:45   that too called rampant layering [TS]

01:22:47   syndrome which was very short little [TS]

01:22:50   thing that basically said but you know [TS]

01:22:52   so these ZFS guys come along and address [TS]

01:22:54   tons of long-standing problems in [TS]

01:22:57   storage all at once in this in this you [TS]

01:22:59   know very technically interesting [TS]

01:23:00   solution and all you can yell them at is [TS]

01:23:02   the fact that they're violating some [TS]

01:23:04   sort of layering and I took it as a [TS]

01:23:07   microcosm of the Linux the Linux world [TS]

01:23:11   at the time this was one was it four [TS]

01:23:13   years ago two thousand seven ish that uh [TS]

01:23:18   in Linux like everyone has their own [TS]

01:23:20   turf and like the file system guys do [TS]

01:23:21   that and device driver guys do that and [TS]

01:23:23   they all want to defend their turf and [TS]

01:23:24   you'll have to have separate layers and [TS]

01:23:26   it makes it so that no one is ever [TS]

01:23:28   thinking about the whole experience it's [TS]

01:23:30   very different from the Apple thing [TS]

01:23:31   where they're like I don't care how [TS]

01:23:32   disgusting it is the whole experience [TS]

01:23:33   has got to be awesome from top to bottom [TS]

01:23:35   and if you have to do some hideous ugly [TS]

01:23:37   stuff in the middle I don't care the [TS]

01:23:39   Apple would much rather have that than a [TS]

01:23:41   product that's mediocre but how these [TS]

01:23:44   layers are so clean and we have a [TS]

01:23:45   division of responsibilities and we the [TS]

01:23:47   layers are changeable independently and [TS]

01:23:49   they say I don't care how good is the [TS]

01:23:50   finished product so I was that got my [TS]

01:23:54   goat a little bit four years ago and [TS]

01:23:57   still bothers me a little bit but you [TS]

01:23:58   know you can see the point about ZFS and [TS]

01:23:59   Microsoft has addressed that with their [TS]

01:24:03   separate storage spaces so they've made [TS]

01:24:04   their decision so I put I put those both [TS]

01:24:06   in the show notes a few more things on r [TS]

01:24:08   EF s a few more details on this they go [TS]

01:24:14   into the description how they didn't [TS]

01:24:15   check something they should have just [TS]

01:24:16   copied and pasted from the ZFS pages [TS]

01:24:18   because so similar to the way they do it [TS]

01:24:19   but some more interesting things I do [TS]

01:24:21   people might be wondering what is it [TS]

01:24:23   that makes a file system you know [TS]

01:24:26   interesting or good when it comes time [TS]

01:24:28   to [TS]

01:24:29   right out all these little pieces of [TS]

01:24:31   information what you usually want to do [TS]

01:24:33   is to keep related information close to [TS]

01:24:35   each other this is especially true in [TS]

01:24:36   the days of discs when you add a little [TS]

01:24:38   disk arm that was moving back and forth [TS]

01:24:39   across the platter [TS]

01:24:40   every time you move that arm it's a huge [TS]

01:24:42   hit in terms of speed so you want to [TS]

01:24:44   just only move to one location and read [TS]

01:24:46   all your stuff so reassess you know does [TS]

01:24:50   what most good file system do is it [TS]

01:24:52   tries to find a way to for very small [TS]

01:24:54   files or for things related to a file [TS]

01:24:56   put everything all in one place so you [TS]

01:24:58   just make it a a one read and you get [TS]

01:24:59   all the information about the file all [TS]

01:25:01   the metadata all the check sums and just [TS]

01:25:03   suck them all in in one nice little [TS]

01:25:05   stream this is another example place [TS]

01:25:07   where our HFS+ does a terrible terrible [TS]

01:25:10   job ah they had the the catalog file I [TS]

01:25:15   think I talked about this in our article [TS]

01:25:16   but the catalog file is the place where [TS]

01:25:18   the HFS+ keeps track of everything and [TS]

01:25:19   there's one of them and it's for the [TS]

01:25:21   entire disc so anytime you want to get [TS]

01:25:23   any information about any file you go to [TS]

01:25:24   go to the catalog file and read it oh [TS]

01:25:25   and by the way if you want the data from [TS]

01:25:27   that file you should go get it elsewhere [TS]

01:25:28   I think they later on started doing some [TS]

01:25:30   stuff where they try to inline stuff [TS]

01:25:31   into the catalog file but in general [TS]

01:25:33   that's a non modern file system design [TS]

01:25:35   modern file systems do not want to have [TS]

01:25:37   a central repository for information [TS]

01:25:39   they always have to go to the central [TS]

01:25:40   place and then go someplace else to get [TS]

01:25:42   the data [TS]

01:25:42   so our EFS is a cognizant of that as as [TS]

01:25:45   is NTFS and the integrity streams thing [TS]

01:25:48   they're actually doing that on a per [TS]

01:25:49   file basis as they imagine that this is [TS]

01:25:52   similar to how they implemented the [TS]

01:25:53   compression thing is that for an [TS]

01:25:54   individual file the the programmer can [TS]

01:25:57   just pass a different flag to their you [TS]

01:25:59   know file out calls and say oh this file [TS]

01:26:01   is going to be you know going to be [TS]

01:26:03   checksum that's how the data checks um [TS]

01:26:05   so that seems like it's something it's [TS]

01:26:06   very least it's very easily accessible [TS]

01:26:08   to the programmer and probably I assume [TS]

01:26:09   will be exposed in some fashion through [TS]

01:26:12   the UI considering how they expose the [TS]

01:26:13   the compression stuff with NTFS so [TS]

01:26:18   finally we'll circle back to the Mac and [TS]

01:26:20   this kind of all came to the head with [TS]

01:26:22   the Rio fest announcement and also with [TS]

01:26:24   the announcement of XIV oze vo from tins [TS]

01:26:28   compliment that is company started by [TS]

01:26:31   Ghana his name room Don Brady tastes the [TS]

01:26:35   things I should have in the show notes [TS]

01:26:36   but don't some annex Apple engineer who [TS]

01:26:39   spent 20 years at Apple working on H of [TS]

01:26:41   s and H of s + [TS]

01:26:43   and was apparently very disappointed [TS]

01:26:46   when Apple decided not to go with ZFS as [TS]

01:26:48   the new file system for the Mac I'm [TS]

01:26:51   guessing he was this point because honey [TS]

01:26:54   there was he left Apple and started his [TS]

01:26:55   own company to bring ZFS to the Mac so [TS]

01:26:58   you don't get much more disappointed [TS]

01:26:59   than that he's got on on the the the [TS]

01:27:02   website it's tends complement calm he's [TS]

01:27:04   got a picture in the first section below [TS]

01:27:07   the giant banner it says out with the [TS]

01:27:09   old and it shows the picture of a 3.5 [TS]

01:27:10   inch floppy disk with the words with the [TS]

01:27:12   letters H FS 1986 written on it and you [TS]

01:27:16   know it's like halt we shouldn't be [TS]

01:27:18   bashing h FS those people did work he [TS]

01:27:20   was probably one of the ones who did [TS]

01:27:21   that work so it's kind of okay for him [TS]

01:27:22   to say look H of s and H of us applause [TS]

01:27:24   were cool I was there when we made them [TS]

01:27:26   I was proud of them at the time but the [TS]

01:27:28   world has moved on and I really think [TS]

01:27:29   the Mac needs a better file system and [TS]

01:27:32   we almost ported ZFS to the Mac got it [TS]

01:27:34   kind of working sort of mostly for the [TS]

01:27:36   napple abandoned in that project so he's [TS]

01:27:38   going to make a commercial product which [TS]

01:27:39   apparently you can do given the license [TS]

01:27:41   that son certainly has that you can buy [TS]

01:27:43   for the Mac that will give you ZFS and [TS]

01:27:46   there's also the Mac ZFS project wasn't [TS]

01:27:49   which is an open source project that has [TS]

01:27:51   a Google Code page that's been ongoing [TS]

01:27:52   for a long time which is hey Apple [TS]

01:27:55   doesn't seem like they're interested in [TS]

01:27:56   this project they've dropped it it's not [TS]

01:27:57   supported anymore let's pick up the [TS]

01:27:59   source that's the magic of open source [TS]

01:28:00   and we'll continue this project so zero [TS]

01:28:02   is a commercial product supported by a [TS]

01:28:04   company it's got a pretty good pedigree [TS]

01:28:07   behind it in terms of if anyone is going [TS]

01:28:09   to understand the intricacies of Mac OS [TS]

01:28:11   10 file system drivers structure it's [TS]

01:28:13   going to be the guy who worked at Apple [TS]

01:28:15   for 20 years working on that stuff right [TS]

01:28:16   so he's got a good shot of doing good [TS]

01:28:18   job and his team and then Mac CFS is the [TS]

01:28:21   typical open source thing is like anyone [TS]

01:28:22   who wants to contribute you know it's a [TS]

01:28:24   free product you get what you pay for so [TS]

01:28:27   these two things are kind of in [TS]

01:28:28   competition with each other that many [TS]

01:28:31   people been asking hi aren't you excited [TS]

01:28:32   about that about the zebu thing ZFS [TS]

01:28:34   finally coming to Mac isn't this awesome [TS]

01:28:36   I totally applaud the idea that you know [TS]

01:28:41   if Apple seems to have its head up its [TS]

01:28:43   butt about file systems and you're the [TS]

01:28:45   file system guy leave the company start [TS]

01:28:48   your own company and sell file system [TS]

01:28:49   for the Mac is open-source Apple can't [TS]

01:28:51   stop you you know what I mean [TS]

01:28:53   that's great and the open source thing [TS]

01:28:55   it if [TS]

01:28:56   want to hack a way out and get to work [TS]

01:28:58   that's great for them too although I [TS]

01:29:00   think they have slightly a deficit of [TS]

01:29:03   knowledge compared to Zeebo if only [TS]

01:29:06   because I suspect there aren't many 20 [TS]

01:29:08   year Apple filesystem veterans working [TS]

01:29:10   on Mac ZFS but in both cases I'm kind of [TS]

01:29:13   sad because a filesystem is especially [TS]

01:29:17   if also smell like ZFS which sort of [TS]

01:29:18   reaches through the IO stack much more [TS]

01:29:20   than most other file systems because of [TS]

01:29:22   you know that violation of the layering [TS]

01:29:25   that people always complain about really [TS]

01:29:28   I would like to come from the OS vendor [TS]

01:29:30   and I get a really kind of nervous about [TS]

01:29:33   file systems that aren't made by the OS [TS]

01:29:35   vendor not through any fault of the [TS]

01:29:37   people making the file systems but just [TS]

01:29:39   because Apple feels no responsibility to [TS]

01:29:43   make sure they don't break the work of [TS]

01:29:44   these people so if Apple changes their [TS]

01:29:49   operating system and the XIV OCFS things [TS]

01:29:51   break Apple isn't shed it's here that [TS]

01:29:53   have you complain and say hey you know I [TS]

01:29:55   upgraded to two mountain lion and I [TS]

01:29:57   can't read my disk or something's messed [TS]

01:29:59   up or they're going to say all right [TS]

01:30:01   let's see what your problem is like oh [TS]

01:30:02   you're using some sort of third-party [TS]

01:30:03   product well that's not our problem [TS]

01:30:04   that's a third parties problem and [TS]

01:30:06   they're kind of right like the OS vendor [TS]

01:30:09   is only supposed to support the volume [TS]

01:30:11   formats that it supports and if you have [TS]

01:30:12   some sort of problem then you have to go [TS]

01:30:14   to the third party but you know how good [TS]

01:30:15   Apple is about keeping third parties and [TS]

01:30:17   loop on things [TS]

01:30:18   it's totally conceivable that Apple [TS]

01:30:20   could release an update to an operating [TS]

01:30:22   system without giving third parties [TS]

01:30:25   either enough time or sometimes any time [TS]

01:30:27   to see whether their things work or not [TS]

01:30:29   like there okay well Zeebo was working [TS]

01:30:30   perfectly fine in the last developer [TS]

01:30:32   seat of this thing but when the final [TS]

01:30:33   version came out we didn't get a chance [TS]

01:30:34   to look at it and always released the [TS]

01:30:35   customers and it broke all their stuff [TS]

01:30:37   so it's really dangerous to be relying [TS]

01:30:41   on a third party for your file system [TS]

01:30:42   needs not through any fault to the third [TS]

01:30:45   party but just because Apple it's so bad [TS]

01:30:46   at you know helping third party Long's [TS]

01:30:49   never these apples position is probably [TS]

01:30:53   we don't think you should use these [TS]

01:30:54   third-party products we have storage [TS]

01:30:56   products we have our own file system [TS]

01:30:57   that's what we support you should use [TS]

01:30:58   that I don't think Apple is going to go [TS]

01:31:01   out of the way to be malicious and break [TS]

01:31:03   stuff but the sure as heck not going to [TS]

01:31:04   care if they break it and file system [TS]

01:31:06   like ZFS I really want to be integrated [TS]

01:31:08   by the platform owner because [TS]

01:31:10   as I want Disk Utility to work I want it [TS]

01:31:12   to be integrated with every part of the [TS]

01:31:15   operating system I want all the apples [TS]

01:31:16   applications to be run and checked on [TS]

01:31:18   ZFS II it really just has to it's not [TS]

01:31:21   that's talking about outsourcing your [TS]

01:31:23   core competency you the platform vendor [TS]

01:31:26   has to support a file system [TS]

01:31:27   soup-to-nuts backups time machine [TS]

01:31:31   everything head to tail so that you know [TS]

01:31:33   you're completely sure that it worked [TS]

01:31:34   before I would say you should trust this [TS]

01:31:37   thing with all of your data [TS]

01:31:38   people are asking chatroom if I have [TS]

01:31:40   tried it I have not no why not try it [TS]

01:31:42   why have I not tried it because I'm [TS]

01:31:44   nervous I'm scared it's scary now here's [TS]

01:31:49   where here's where Zeebo is going to [TS]

01:31:50   find an audience if you have a whole [TS]

01:31:52   bunch of disks that you want to first of [TS]

01:31:54   all I don't think you can even boot from [TS]

01:31:55   it yet so it's obviously out of the [TS]

01:31:57   realm of possibility or interests for [TS]

01:31:59   most people because most people have one [TS]

01:32:00   disk and maybe a time machine disk it's [TS]

01:32:04   out of the question camp boot but if you [TS]

01:32:08   have a whole bunch of disks that you [TS]

01:32:09   want to like store a bunch of stuff on [TS]

01:32:11   like your home Nass or you have a little [TS]

01:32:13   Mac Mini attached to and you want to [TS]

01:32:14   make a whole nest the type of thing [TS]

01:32:16   that's where I would start looking into [TS]

01:32:18   this because then you're like well I'm [TS]

01:32:19   going to build my own Drobo type of [TS]

01:32:21   thing right ah and that's something I [TS]

01:32:24   want to do because I can do it much more [TS]

01:32:25   cheaply in a drawer because I have this [TS]

01:32:26   old Mac Mini hanging around and I could [TS]

01:32:27   just attach a bunch of these drives off [TS]

01:32:29   with firewire what can I do with these [TS]

01:32:31   you know three or four drives attached [TS]

01:32:32   with firewire to a Mac Mini well if you [TS]

01:32:35   use Apple and HFS+ you can use their [TS]

01:32:37   software great which is terrible [TS]

01:32:38   hardware RAID is out of the question you [TS]

01:32:41   could buy a Drobo box but those cost a [TS]

01:32:42   lot of money on top of the disk you have [TS]

01:32:43   to do is like I've just got the disks [TS]

01:32:45   and I've got a Mac Mini what can I do [TS]

01:32:46   z-bo is the answer there you can SiC it [TS]

01:32:49   at those disks combine them in together [TS]

01:32:51   one big storage pool and you'll have all [TS]

01:32:53   the benefits of ZFS with the you know [TS]

01:32:55   checksumming and redundancy and the [TS]

01:32:57   atomic updates and all that good stuff [TS]

01:32:59   and if you just use it to store like [TS]

01:33:01   like a home video server or something [TS]

01:33:03   and you back it up using some other [TS]

01:33:05   mechanism you know are you backing up [TS]

01:33:07   using another one of those things then [TS]

01:33:08   you're good to go it's just a shame that [TS]

01:33:11   you know I want that check something and [TS]

01:33:14   all that stuff and all the performance [TS]

01:33:15   benefits and stuff to be from my main [TS]

01:33:18   disk for the one that has my operating [TS]

01:33:20   system or the one that has my [TS]

01:33:21   application like the one that has my [TS]

01:33:22   iPhone [TS]

01:33:23   but at this point I'm you know I don't [TS]

01:33:27   think for me personally anyway zero is [TS]

01:33:29   not the solution to that maybe Apple [TS]

01:33:33   will buy Zeebo and fold them back in and [TS]

01:33:34   incorporate it but I resent seem like [TS]

01:33:36   late to you I don't know well you say [TS]

01:33:38   this is this gets to the root of this [TS]

01:33:40   thing so I did all this complaining we [TS]

01:33:42   talked about all file systems we talked [TS]

01:33:44   about how they didn't get ZFS what ZFS [TS]

01:33:46   and we talked about how Microsoft has [TS]

01:33:50   decided to up its file system game with [TS]

01:33:52   three FS what's what's apples move what [TS]

01:33:55   are they doing that's that's the big [TS]

01:33:57   question and the answer seems so far has [TS]

01:34:00   been nothing everything's fine more or [TS]

01:34:02   less well add feature state service plus [TS]

01:34:04   when we need to we need compression we [TS]

01:34:06   need to add a logical volume manager so [TS]

01:34:08   we can do whole disk encryption well we [TS]

01:34:10   need to do stuff we'll do it [TS]

01:34:11   but Ava's plus is fine more or less and [TS]

01:34:13   I don't that's an answer I don't like [TS]

01:34:16   but so how long does that go on 10 years [TS]

01:34:18   20 years 50 years am I going to die and [TS]

01:34:21   the default file system for Macs or [TS]

01:34:23   something Mac's are still around is HFS+ [TS]

01:34:26   yep the the world has moved on so far [TS]

01:34:29   since you know from between the time [TS]

01:34:31   HFS+ was introduced in now everyone else [TS]

01:34:35   is moving on everyone else is doing [TS]

01:34:36   something different even Microsoft is [TS]

01:34:38   doing something different goals got its [TS]

01:34:39   own file system the data center [TS]

01:34:40   certainly it's not using like fat disks [TS]

01:34:42   or anything like that it's got GFS going [TS]

01:34:44   on over there [TS]

01:34:45   hey I this is something that Apple needs [TS]

01:34:49   to address and so far they haven't so [TS]

01:34:52   this is the great mystery of of Apple [TS]

01:34:55   what why'd you know why don't they [TS]

01:34:56   address the possum thing it clearly I [TS]

01:34:59   think there's something to be addressed [TS]

01:35:00   because they did that whole ZFS fiasco [TS]

01:35:02   right you don't do that just for your [TS]

01:35:04   health isn't like they thought that HFS+ [TS]

01:35:07   should be replaced with something better [TS]

01:35:09   and they had this big debate or the [TS]

01:35:10   better thing should be and they kind of [TS]

01:35:12   pick ZFS then it fell through well don't [TS]

01:35:13   they still think exactly that the David [TS]

01:35:15   HMS plays needs to be replaced like it [TS]

01:35:17   and you know if anything I think the [TS]

01:35:19   people who are working on HFS+ at Apple [TS]

01:35:21   would be the most vocal about the fact [TS]

01:35:23   that geez we can't keep bolting crap to [TS]

01:35:25   this thing like you if you worked on any [TS]

01:35:26   piece of software for a long period of [TS]

01:35:28   time you know you can improve it and you [TS]

01:35:29   can refactor it but the fundamental [TS]

01:35:31   design decisions of HLS plus are not [TS]

01:35:33   appropriate for the current age having a [TS]

01:35:35   central catalog by [TS]

01:35:36   single-threaded accesses and things I [TS]

01:35:38   talked about line article we need to [TS]

01:35:39   mention that single thread access only [TS]

01:35:40   one only one the process can be [TS]

01:35:42   accessing the file system at once [TS]

01:35:43   doesn't that sound like something for [TS]

01:35:44   the monitor of 16 core CPUs no so sound [TS]

01:35:48   likely the ideal implementation of a [TS]

01:35:50   disk the made up of multiple spindles in [TS]

01:35:52   a volume it up of multiple disks it's [TS]

01:35:55   it's our kick so Apple needs to do [TS]

01:35:58   something and I don't at this point and [TS]

01:35:59   at previous points I don't really care [TS]

01:36:01   what that thing is but please just do [TS]

01:36:02   something but make a new file system [TS]

01:36:04   yourself adopt btrfs adopts EFS by Zeebo [TS]

01:36:08   so we talked about what is the [TS]

01:36:10   possibility of them buying out Zeebo [TS]

01:36:13   maybe the things that it it all depends [TS]

01:36:16   on what it is that made them not go with [TS]

01:36:18   ZFS it's not like it's not like Apple's [TS]

01:36:20   forthcoming in that regard they took it [TS]

01:36:21   over the webpage and I said yeah we're [TS]

01:36:23   not doing that then there was no big [TS]

01:36:24   long heart-to-heart about why they why [TS]

01:36:26   they'd made the decision so everyone's [TS]

01:36:27   speculating but it was because of [TS]

01:36:29   licensing with with Oracle or Sun or if [TS]

01:36:33   it was because of legal concerns with [TS]

01:36:34   NetApp I think all that stuff still [TS]

01:36:35   applies because a licensing issue hasn't [TS]

01:36:37   changed as far as I know and the legal [TS]

01:36:39   thing is still pending so if that's what [TS]

01:36:41   scared them away that will continue to [TS]

01:36:43   scare them away they're they're not [TS]

01:36:44   going to buy 0z but if those are tends [TS]

01:36:47   complement rather but if that if that [TS]

01:36:50   stuff is not applicable if it's been [TS]

01:36:52   resolved or wasn't the reason in the [TS]

01:36:53   first place or whatever and those Evo [TS]

01:36:55   guys do a bang-up job and you know that [TS]

01:36:58   that would be an ideal acquisition [TS]

01:37:00   because otherwise what are you gonna do [TS]

01:37:02   Apple we don't want to buy Zeebo we [TS]

01:37:04   don't want to do ZFS well what do you [TS]

01:37:06   want to do like he got to do something [TS]

01:37:08   eventually that's why I think it's got [TS]

01:37:09   to come to a head I keep saying this [TS]

01:37:10   like I said with it with the Copeland [TS]

01:37:11   2010 business about you gotta do [TS]

01:37:13   something about your language eventually [TS]

01:37:14   and so they keep just tacking stuff on [TS]

01:37:16   to Objective C well you do something on [TS]

01:37:18   your file system eventually they keep [TS]

01:37:19   tacking stuff on to HFS+ it it's going [TS]

01:37:21   to come to a head and and unlike the [TS]

01:37:23   Copeland 2010 stuff it is much easier to [TS]

01:37:26   change your default file system or add [TS]

01:37:28   support for new file system and slowly [TS]

01:37:29   transition everybody is than it is to [TS]

01:37:31   change your language or API it seems [TS]

01:37:35   easier to me not that I'm saying it's [TS]

01:37:36   easy it's going to be hard you have to [TS]

01:37:37   do it carefully but they've done it at [TS]

01:37:39   least twice before three times they went [TS]

01:37:41   from a manifest stage of us they went [TS]

01:37:42   from HR estivate to s+ and arguably they [TS]

01:37:45   went from hfs+ to 8 billion variants of [TS]

01:37:48   HFS+ with journaling [TS]

01:37:49   with transparent encryption and then [TS]

01:37:51   course storage so if Apple has a plan [TS]

01:37:54   they're not saying but I they definitely [TS]

01:37:56   need to do something I think it's a lot [TS]

01:38:00   of got on file systems unless you think [TS]

01:38:02   there's something I missed like one I [TS]

01:38:09   know I think that's it it's a lot yeah [TS]

01:38:17   it's it's one of the eternal mysteries [TS]

01:38:19   of Apple you know people talk about they [TS]

01:38:20   made the phone they made a tablet Macs [TS]

01:38:23   are on Intel like what's you know what's [TS]

01:38:26   the mystery I guess you got the TV still [TS]

01:38:28   it's things so gone but they kind of you [TS]

01:38:29   know people still say that's the mystery [TS]

01:38:31   but they did Apple TV wanted to the [TS]

01:38:33   Apple TV - and we're suppose you know [TS]

01:38:34   what they can do for TV they did two [TS]

01:38:36   things already so this may be one of the [TS]

01:38:38   remaining mysteries of Apple what you [TS]

01:38:41   can do about the file system Apple it's [TS]

01:38:42   kind of boring mystery that no one is [TS]

01:38:44   really interested in but I am and the [TS]

01:38:47   answer is we don't know I mean but isn't [TS]

01:38:49   isn't the whole goal inevitably to take [TS]

01:38:52   the concept the entire concept of the [TS]

01:38:56   file system off the plate for users so [TS]

01:39:01   that we don't you know I mean this is [TS]

01:39:03   it's always silly to go back to make a [TS]

01:39:05   car analogy but but it's sometimes it [TS]

01:39:08   works most people you you probably know [TS]

01:39:12   exactly what kind of engine you have in [TS]

01:39:14   your car and whether it's you know an [TS]

01:39:16   older one that's carbureted or [TS]

01:39:18   fuel-injected you know all the probably [TS]

01:39:20   things you know how many horsepower [TS]

01:39:21   everything a car aficionados people who [TS]

01:39:24   are into that mechanics people who like [TS]

01:39:26   cars and get into it they know all of [TS]

01:39:28   those details and that's fine but there [TS]

01:39:33   are a whole lot of people out there who [TS]

01:39:36   don't know what kind of engine or if [TS]

01:39:39   their car even has an engine they know [TS]

01:39:41   they turn the key and they know how to [TS]

01:39:43   operate the vehicle and sometimes if it [TS]

01:39:45   makes noise or doesn't start they [TS]

01:39:47   somebody else comes out and fixes it [TS]

01:39:49   takes it away and they pay money and it [TS]

01:39:51   comes back and it works again the fact [TS]

01:39:54   that they know that there is an engine [TS]

01:39:55   in there they know that because they've [TS]

01:39:57   read that they've been told that when [TS]

01:39:59   they turn the key the car makes noise [TS]

01:40:01   they put gas they understand the [TS]

01:40:02   Sept of it but you could say wouldn't it [TS]

01:40:06   wouldn't it be a beneficial thing at [TS]

01:40:07   some point if maybe this is fifty or a [TS]

01:40:10   thousand years down the road who knows [TS]

01:40:12   but that at some point we'll have a [TS]

01:40:14   vehicle it won't make any sound it won't [TS]

01:40:16   have any emissions you'll just push a [TS]

01:40:18   button you'll say where you want to go [TS]

01:40:20   and you'll go there the thing will take [TS]

01:40:21   you there you won't know how it works [TS]

01:40:23   and it won't be important to know how it [TS]

01:40:25   works sure some people will know how it [TS]

01:40:27   works but the general people they'll [TS]

01:40:29   have no idea most people don't know how [TS]

01:40:30   electricity gets into there into the [TS]

01:40:32   socket and they plug their computer into [TS]

01:40:34   they just know that there's there's [TS]

01:40:36   power in there somehow it's just reading [TS]

01:40:39   you're stretching your analogy too far [TS]

01:40:41   well I liked where you were going with [TS]

01:40:43   the car we'll take a step back and we'll [TS]

01:40:45   say that shouldn't apple's goal be to [TS]

01:40:49   give to take the word filesystem out of [TS]

01:40:52   our vocabulary and just know that there [TS]

01:40:55   are these things that you make and [TS]

01:40:57   they're there in this place and wherever [TS]

01:41:02   you are you can get to them and have [TS]

01:41:03   them I mean that's the goal right that's [TS]

01:41:05   what Apple wants to do so take away the [TS]

01:41:07   filesystem did you know thinking of [TS]

01:41:09   iCloud as the place where everything [TS]

01:41:11   lives I don't even know how the data got [TS]

01:41:15   on this device or this computer or where [TS]

01:41:17   I created it it's irrelevant I can [TS]

01:41:19   access it here on this device I happen [TS]

01:41:21   to be in front of right now I'll get all [TS]

01:41:23   the stuff that I want right here I make [TS]

01:41:25   a change it's everywhere and I is there [TS]

01:41:27   a filesystem oh I don't know my iOS [TS]

01:41:29   device just died but I don't have to [TS]

01:41:31   worry about the contents of the things [TS]

01:41:33   on that iOS device because they're in [TS]

01:41:35   that place the magical place where all [TS]

01:41:37   of my data lives and all all these [TS]

01:41:40   devices and computers and things are [TS]

01:41:42   simply just a portal I don't know how it [TS]

01:41:44   works I don't care how it works I don't [TS]

01:41:46   know if there's ZFS if there's raid I [TS]

01:41:48   don't know who XIV o is or what kind of [TS]

01:41:50   music he plays it's irrelevant all I all [TS]

01:41:53   I know is that I my data is in this [TS]

01:41:55   magical place and I can get to it [TS]

01:41:58   wherever I am with any device I have [TS]

01:41:59   that's the Apple vision right so I think [TS]

01:42:02   you're bringing up a good point that I [TS]

01:42:04   should have had in my notes which is [TS]

01:42:05   when I already talked about file systems [TS]

01:42:07   somebody usually someone kind of nerdy [TS]

01:42:11   either wants to talk about or thinks I'm [TS]

01:42:14   talking about [TS]

01:42:16   the idea that you of the what would you [TS]

01:42:20   call it the conceptual structure created [TS]

01:42:26   by the file system that it's at the top [TS]

01:42:29   level you know like it just a UNIX thing [TS]

01:42:31   starts at slash and there's a bunch of [TS]

01:42:32   folders inside those other stuff other [TS]

01:42:34   files and other folders it's a big tree [TS]

01:42:35   of stuff and to locate anything you have [TS]

01:42:37   to know where it is in this big [TS]

01:42:38   hierarchy and you go get it and when you [TS]

01:42:42   talk about getting rid of the file [TS]

01:42:43   system or something like that well [TS]

01:42:44   people are mostly talking about is stop [TS]

01:42:46   making people locate things by hunting [TS]

01:42:49   through folders that are nested inside [TS]

01:42:50   each other [TS]

01:42:50   stop making people be aware of file [TS]

01:42:52   paths don't you know kind of like how I [TS]

01:42:55   voted it of your photos you don't know [TS]

01:42:56   where they are and if you're not [TS]

01:42:57   supposed to know they're managed by the [TS]

01:42:58   application it's simplifying that vision [TS]

01:43:01   because as we all know for many years [TS]

01:43:02   experience in dealing with computers [TS]

01:43:03   most people do not understand [TS]

01:43:05   quote-unquote the file system and when [TS]

01:43:07   we say that what we mean is people don't [TS]

01:43:08   know where the heck things are they [TS]

01:43:09   don't they don't visualize the hierarchy [TS]

01:43:11   of folders that's rep that's created by [TS]

01:43:14   you know the illusion of the hierarchy [TS]

01:43:15   Foles is create about the file system [TS]

01:43:17   they'd have it's not in their head so [TS]

01:43:19   they don't know where things are there [TS]

01:43:21   are many reasons for that but I go into [TS]

01:43:22   my whole big spatial find a rant again [TS]

01:43:24   but the bottom line is they don't [TS]

01:43:25   understand it they don't understand open [TS]

01:43:27   save dialog boxes the only place they [TS]

01:43:28   know how to find is the desktop because [TS]

01:43:30   it's underneath everything anything [TS]

01:43:32   beyond that they have no idea and it [TS]

01:43:34   just it's it's not a good fit for people [TS]

01:43:36   and as I think I said at some point [TS]

01:43:39   maybe was one of these shows maybe in a [TS]

01:43:40   review at this point in the experiment I [TS]

01:43:42   think we can confidently say that it's [TS]

01:43:44   not a problem of Education there are [TS]

01:43:46   generations of people who are grown up [TS]

01:43:47   with computers their entire lives and [TS]

01:43:49   still have no idea where that anything [TS]

01:43:51   is obviously that way of organizing [TS]

01:43:53   files while it works great for for our [TS]

01:43:55   nerds computer geeks does not work great [TS]

01:43:57   for most people and so it's a bad model [TS]

01:44:00   so Apple true to form is going to say [TS]

01:44:01   we're gonna solve this problem by let's [TS]

01:44:05   take away that assumption let's let's [TS]

01:44:07   not we don't know what we're going to do [TS]

01:44:08   instead but we don't want to make you [TS]

01:44:10   hunt around for stuff so iTunes it [TS]

01:44:12   manages all your stuff you use not [TS]

01:44:13   supposed to care whether they are on [TS]

01:44:14   disk iPhoto it manages all your photos [TS]

01:44:15   and I suppose to have folders full of [TS]

01:44:17   stuff that you that you're manually [TS]

01:44:18   renamed in two years and do all this [TS]

01:44:19   stuff and we have to keep track of it [TS]

01:44:21   we're just gonna hide that from right [TS]

01:44:23   that's not what I'm talking about when I [TS]

01:44:25   talk about file systems I know the word [TS]

01:44:26   is overloaded what I'm talking about in [TS]

01:44:28   this goes back to your car engine [TS]

01:44:29   gee which is a good one is that you know [TS]

01:44:32   people don't know what's inside the the [TS]

01:44:34   under the hood of their car they don't [TS]

01:44:35   care they just want the thing to go [TS]

01:44:36   that's all true but the Apple situation [TS]

01:44:39   would be as if Apple the car maker was [TS]

01:44:42   still selling carbureted engines if you [TS]

01:44:45   know ignoring regulations and and all [TS]

01:44:47   those you know emissions and stuff like [TS]

01:44:48   that they're like oh you don't have to [TS]

01:44:50   know what's in the cover just by under [TS]

01:44:51   the hood just buy what you want an Apple [TS]

01:44:53   engine and it was like an engine from [TS]

01:44:54   1972 the big giant air filter and a [TS]

01:44:57   carburetor on it and you know [TS]

01:44:59   tremendously low horsepower per liter [TS]

01:45:01   and it sucked gas like crazy and it spit [TS]

01:45:04   out let you know even though perhaps [TS]

01:45:07   especially since modern car has been a [TS]

01:45:09   less and less about what's under the [TS]

01:45:10   hood engine technology is advanced [TS]

01:45:13   tremendously the number of the amount of [TS]

01:45:15   power we get out of a liter displacement [TS]

01:45:16   has been increasing tremendously fuel [TS]

01:45:19   efficiency is going up we have cylinder [TS]

01:45:20   deactivation we have increases advances [TS]

01:45:23   in the materials used to make engines it [TS]

01:45:26   advances in turbo charging or moving [TS]

01:45:28   turbo lag and superchargers and you know [TS]

01:45:30   hybrid technology engine technology [TS]

01:45:32   advances like crazy and there's not a [TS]

01:45:33   single car manufacturer that can get [TS]

01:45:35   away with saying no engines good enough [TS]

01:45:37   that's all right [TS]

01:45:39   oh we don't need any of that stuff we [TS]

01:45:40   don't need any of those friction [TS]

01:45:41   reducing things we don't need overhead [TS]

01:45:43   cams we need direct injection cylinder [TS]

01:45:46   deactivation that seems pointless and [TS]

01:45:47   our engine is fine don't worry about it [TS]

01:45:49   it's under the hood the car will go [TS]

01:45:50   it'll be fine engine is fine you're not [TS]

01:45:52   to worry about and you know in car the [TS]

01:45:54   car world you'd say well certainly it's [TS]

01:45:56   going to come to head where someone's [TS]

01:45:57   you know again ignoring government [TS]

01:45:58   regulation like I say I get 7 miles a [TS]

01:46:00   gallon out of this car I'm not gonna buy [TS]

01:46:02   any more of these Apple cars they get 7 [TS]

01:46:04   miles again I'm gonna buy a Honda you [TS]

01:46:05   got 37 miles to the gallon I'm gonna buy [TS]

01:46:07   a hybrid and get 50 or you know whatever [TS]

01:46:09   you can do eventually consumers start to [TS]

01:46:12   notice the difference so far Apple has [TS]

01:46:14   been betting that with these little [TS]

01:46:15   tweaks that it's making to its ancient [TS]

01:46:17   you know catalytic converter not get [TS]

01:46:21   like a rotor carburetor sporting engine [TS]

01:46:23   under the hood that people won't notice [TS]

01:46:25   those thing that's good enough [TS]

01:46:27   especially since they don't compare it [TS]

01:46:29   to usually because max only run the [TS]

01:46:31   nature best boss you can't install Mac [TS]

01:46:33   os10 on some other process and then [TS]

01:46:34   compare the performance right so you [TS]

01:46:37   can't say whoa I installed [TS]

01:46:38   you know I'm booting Mac os10 from a ZFS [TS]

01:46:41   volume and it's way faster and [TS]

01:46:43   incidentally and maybe the case ZFS [TS]

01:46:45   isn't faster depending on how the [TS]

01:46:46   implementation deals with a check sums [TS]

01:46:47   and everything but any any other file [TS]

01:46:48   system you can't do a comparison and [TS]

01:46:51   since computers and harbours keep [TS]

01:46:53   getting faster your Mac keeps getting [TS]

01:46:55   faster too and it's kind of it's harder [TS]

01:46:56   for consumers to notice you know I think [TS]

01:46:59   this will be faster if if Apple's file [TS]

01:47:01   system coalesced iOS into larger blocks [TS]

01:47:04   and deferred them so it could do them [TS]

01:47:05   all at once so I think this would be [TS]

01:47:06   faster if more than one thread could [TS]

01:47:08   access the fauces no one sees your [TS]

01:47:09   thoughts that are not entering someone's [TS]

01:47:10   head but for people who know about car [TS]

01:47:12   engines or file systems we can all see [TS]

01:47:14   yeah did this horrible smoke belching [TS]

01:47:16   thing that Apple's got under the hood [TS]

01:47:17   is so far behind the times and we just [TS]

01:47:19   think boy imagine how much better it [TS]

01:47:21   could be if we had a modern high [TS]

01:47:23   performance entry and you know naturally [TS]

01:47:25   aspirated getting a hundred and ten [TS]

01:47:26   horsepower per liter out of it with [TS]

01:47:28   titanium connecting rods and just you [TS]

01:47:30   know we know what's possible because we [TS]

01:47:32   see everyone else doing it so there are [TS]

01:47:35   two totally independent things of [TS]

01:47:36   navigating the hierarchy and the idea of [TS]

01:47:38   that whole structure and look there's [TS]

01:47:41   got to be something to store the data [TS]

01:47:42   locally under the hood whatever that [TS]

01:47:44   thing is make it better faster stronger [TS]

01:47:46   protect your data all that stuff and and [TS]

01:47:48   again for the like I don't care if it's [TS]

01:47:49   here I don't care if my devices bad [TS]

01:47:51   always get it from the cloud again uh [TS]

01:47:53   maybe that will be their solution that [TS]

01:47:55   eventually actually I don't think that [TS]

01:47:57   can be their solution either because you [TS]

01:47:59   say oh well if I drop my thing in the [TS]

01:48:00   toilet all my data is in the cloud and I [TS]

01:48:02   don't worry about on the cloud and [TS]

01:48:03   Apple's not putting their filesystem on [TS]

01:48:04   a cloud they're using some storage from [TS]

01:48:05   you know EMC or whatever behind the [TS]

01:48:07   scenes and that's all enterprise-e [TS]

01:48:08   storage and cost tons of money and it [TS]

01:48:10   protects your data and blah blah blah I [TS]

01:48:12   don't have to care right well you do [TS]

01:48:13   have the care because if they used in [TS]

01:48:15   some crappy file system on your iPhone [TS]

01:48:17   or whatever and it's stuff gets [TS]

01:48:19   corrupted and it doesn't notice and it [TS]

01:48:20   uploads the crappy corrupted stuff to [TS]

01:48:22   the cloud the next time you get a device [TS]

01:48:23   and download it you're going to get [TS]

01:48:24   corrupted stuff right back that the [TS]

01:48:26   cloud dutifully preserved for you and if [TS]

01:48:28   your thing has no idea that it's [TS]

01:48:30   corrupted no one's ever going to notice [TS]

01:48:31   because the cloud says this is the data [TS]

01:48:33   I got it's not corrupted or I'm [TS]

01:48:34   concerned I'm checksumming and it's [TS]

01:48:35   exactly the data I got but the phone [TS]

01:48:37   corrupted didn't had know it to tell or [TS]

01:48:39   your Mac corrupted it had no way to tell [TS]

01:48:41   so this this is a problem that [TS]

01:48:42   absolutely must be solved eventually and [TS]

01:48:44   all we're arguing about is when that [TS]

01:48:46   eventually comes but we're done arguing [TS]

01:48:51   Rp [TS]

01:48:54   you feel calm confident but you covered [TS]

01:48:58   it all [TS]

01:48:59   we'll see you next week I predict you [TS]

01:49:02   will have some a lot of follow-up here [TS]

01:49:04   blah I don't know how interested people [TS]

01:49:05   really are in filesystems because of the [TS]

01:49:07   people who are real files and nerds I [TS]

01:49:09   didn't say anything they already didn't [TS]

01:49:10   know and the people who aren't file [TS]

01:49:11   system Nerds may have been bored of [TS]

01:49:12   their mind but we'll see who is not a [TS]

01:49:15   filesystem nerd so many people so many [TS]

01:49:19   let me ask you this would you rather it [TS]

01:49:21   just go away [TS]

01:49:22   would you rather to cut the concept yes [TS]

01:49:24   of us personally no cuz I'm one of the [TS]

01:49:29   people who understands the filesystem [TS]

01:49:30   hierarchy I like it when applications [TS]

01:49:32   take over for me I like the iPhoto - [TS]

01:49:35   lets me not know where stuff is I let [TS]

01:49:36   iTunes organize my music folders but [TS]

01:49:39   anyone who's a developer for a living [TS]

01:49:41   like realizes how far we are from the [TS]

01:49:43   very least from developers getting away [TS]

01:49:45   from that because yeah everything in [TS]

01:49:47   development is based off file paths that [TS]

01:49:49   you are aware of things have to be in [TS]

01:49:51   certain locations to work correctly [TS]

01:49:52   there things are certain well-known [TS]

01:49:54   locations where stuff is stored or you [TS]

01:49:55   have to set paths you know for me no I [TS]

01:49:58   wouldn't like to go it but for consumer [TS]

01:50:00   applications like iTunes and iPhoto I [TS]

01:50:02   like that but there is a lot a long way [TS]

01:50:08   to go to get good solution because you [TS]

01:50:10   know we all we've talked about those [TS]

01:50:12   things before where it's like well [TS]

01:50:13   iPhoto manages your photos or your iOS [TS]

01:50:14   App Store there are their documents but [TS]

01:50:16   how do you share them and is this [TS]

01:50:17   tension between the old way and the new [TS]

01:50:19   way and it's like well there's [TS]

01:50:20   advantages in the old way in terms of [TS]

01:50:21   having a single file worked on by [TS]

01:50:23   multiple applications but once you get [TS]

01:50:24   on that path then how do you find the [TS]

01:50:26   file and what if you have too many files [TS]

01:50:27   and apples working on that stuff I have [TS]

01:50:30   more confidence in Apple's ability to [TS]

01:50:32   experiment and figure out what to do in [TS]

01:50:34   terms of hiding the file hierarchy let's [TS]

01:50:37   call it so I don't get confused from the [TS]

01:50:38   user I don't think they've got that [TS]

01:50:40   right yet and I don't ever want it to be [TS]

01:50:42   hidden from me but I'm a nerd but I [TS]

01:50:43   think Apple will figure that out but the [TS]

01:50:45   under the covers what's under the hood [TS]

01:50:47   of your car they need to make that a [TS]

01:50:49   modern engine and that doesn't matter [TS]

01:50:51   what they do with that list they need a [TS]

01:50:53   modern engine there for something it's [TS]

01:50:55   got to be stored somewhere you got to [TS]

01:50:56   store something somewhere they want to [TS]

01:50:57   make a files and that has no hierarchy I [TS]

01:50:58   don't care but make it reliable make it [TS]

01:51:00   fast make it checksum you know make it [TS]

01:51:03   high-performance all those modern things [TS]

01:51:05   that all the other [TS]

01:51:05   eyes are doing Apple needs to do somehow [TS]

01:51:13   alright then the end the email will be [TS]

01:51:17   coming in if you want to send the email [TS]

01:51:18   you go to five by five dot TV slash [TS]

01:51:20   contact and you pick hypercritical from [TS]

01:51:24   that you can fill that out and we'll [TS]

01:51:25   send it to John and I and or me and John [TS]

01:51:29   as he likes to say we will be able to [TS]

01:51:33   read it perhaps comment it perhaps write [TS]

01:51:35   it on the error respond to it John [TS]

01:51:37   you're very good about that and you care [TS]

01:51:38   I don't I don't see see you on the [TS]

01:51:41   replies though so you don't even know [TS]

01:51:42   whoever blind to I don't but I know that [TS]

01:51:44   you respond to a lot of them and I know [TS]

01:51:47   that in some cases you will be prompted [TS]

01:51:49   by an email to devote a segment of the [TS]

01:51:52   show to respond to responding in the [TS]

01:51:55   only promise that make is that I will [TS]

01:51:57   read them all and I do you've said it [TS]

01:52:01   before in other shows that you don't [TS]

01:52:02   feel that just because somebody sends [TS]

01:52:03   you an email [TS]

01:52:04   you don't feel obligated to respond to [TS]

01:52:07   it even though even to read it but [TS]

01:52:09   certainly not to respond to it no I [TS]

01:52:10   always feel obligated to read it I read [TS]

01:52:13   every email again unless it's spam or [TS]

01:52:15   something right but respond no and [TS]

01:52:17   unsolicited PR email which I get a lot [TS]

01:52:20   of I don't feel an obligation to read [TS]

01:52:21   that but people who are emailing me [TS]

01:52:23   about the show or personally or whatever [TS]

01:52:24   I read all the email do not respond to [TS]

01:52:26   all hmmm people who do that that's a [TS]

01:52:29   whole other show we should do is I'm [TS]

01:52:30   back to work think that I would love to [TS]

01:52:32   get one of those people on the show and [TS]

01:52:33   say what makes you do that people you do [TS]

01:52:37   know people like that who respond every [TS]

01:52:38   email they get I I try to respond to all [TS]

01:52:41   of them but I'm very I'm very bad at it [TS]

01:52:44   some beginning to take your approach [TS]

01:52:45   which is to just not respond at all but [TS]

01:52:48   I feel bad because there are people who [TS]

01:52:50   are asking questions and then they would [TS]

01:52:52   you know they need help or they're they [TS]

01:52:53   want to you know they you read their [TS]

01:52:57   email they tell a story I'll tell you [TS]

01:52:59   what they may I don't know if this is [TS]

01:53:02   true for you or not but I have I have [TS]

01:53:04   some tips that I've read from other [TS]

01:53:06   people who get probably way more email [TS]

01:53:09   than than I do I've only got maybe 3,000 [TS]

01:53:12   emails to respond to in my inbox right [TS]

01:53:14   now I'm not exaggerating but I know [TS]

01:53:17   people who get way way way way more than [TS]

01:53:20   that and the advice that I've often [TS]

01:53:25   heard is you know how they have that and [TS]

01:53:28   I know you're a big fan of the TL [TS]

01:53:31   semi-colon dr concept which is in one's [TS]

01:53:37   have a descriptive subject if you if [TS]

01:53:39   you're putting in a subject but if [TS]

01:53:40   you're not don't worry about and have [TS]

01:53:42   the first line of email be one sentence [TS]

01:53:45   long that expresses everything else and [TS]

01:53:47   I get a lot of emails from people and [TS]

01:53:50   and this is why I feel bad about it [TS]

01:53:51   there are people who write in for [TS]

01:53:53   example about this show they're they [TS]

01:53:54   know how much they loved the show you [TS]

01:53:56   know they this is their favorite episode [TS]

01:53:58   this is something that they really liked [TS]

01:54:00   here's the thing they disagree and after [TS]

01:54:02   five or six paragraphs in the actual [TS]

01:54:05   reason that they wrote is here's a neat [TS]

01:54:09   link showing what you said was true or [TS]

01:54:12   false or back up information so put that [TS]

01:54:14   as the first sentence um you know if [TS]

01:54:18   it's a link say here's an article about [TS]

01:54:20   this relating to episode this that's the [TS]

01:54:23   and if you have more to say say it later [TS]

01:54:24   but it's so hard to read you know a four [TS]

01:54:28   or five six paragraph email they [TS]

01:54:30   obviously somebody spent a lot of time [TS]

01:54:32   writing and they care and they really [TS]

01:54:33   want us to read it but like a lot of the [TS]

01:54:37   emails that we get are I don't know it's [TS]

01:54:39   so I feel I do feel some obligation to [TS]

01:54:42   if somebody spent time writing an email [TS]

01:54:44   I want to I want to spend time reading [TS]

01:54:47   it and if I'm there then yeah I feel [TS]

01:54:49   like I should reply to it John you don't [TS]

01:54:51   think you don't think yeah I know you [TS]

01:54:53   say you don't owe it to anybody but [TS]

01:54:54   don't [TS]

01:54:55   isn't there some kind of conscience in [TS]

01:54:57   there or some kind of sense of [TS]

01:54:58   obligation to respond I I just don't [TS]

01:55:04   feel that because they did a certain [TS]

01:55:05   point like here's but the way I view it [TS]

01:55:07   is it's an it is not logical right if I [TS]

01:55:11   respond to this person if people ask [TS]

01:55:13   questions here's my thing it's a [TS]

01:55:14   cost-benefit analysis if someone asks a [TS]

01:55:16   simple question I respond to it and give [TS]

01:55:17   them an answer it was like it takes two [TS]

01:55:19   seconds to do that you know so I'll just [TS]

01:55:21   do it right ah but some emails don't [TS]

01:55:25   have a clear question in them or are [TS]

01:55:27   more of a statement with an implied [TS]

01:55:29   question or just really long and asked [TS]

01:55:31   tons and tons of questions [TS]

01:55:32   you know what I mean and so I think if I [TS]

01:55:35   were to respond to this person first if [TS]

01:55:37   I think I can actually help them and I [TS]

01:55:39   were to respond to them only that person [TS]

01:55:41   gets the benefit or as if I discuss it [TS]

01:55:43   on the snow on the show everybody who [TS]

01:55:44   listens gets the benefit of ah well [TS]

01:55:47   these 1/2 of the exchange are me [TS]

01:55:49   providing what I think is the answer [TS]

01:55:50   right so I'm thinking like it I don't [TS]

01:55:52   like to expend a lot of effort in a long [TS]

01:55:55   email to a single person when I can get [TS]

01:55:59   more bang for my buck by saying that [TS]

01:56:01   same thing on the show is so many people [TS]

01:56:03   here so that factors into it too right [TS]

01:56:04   but even beyond that after a certain [TS]

01:56:06   volume of email uh it's just not [TS]

01:56:09   possible to respond to everybody it's [TS]

01:56:11   just not I mean this the feedback from [TS]

01:56:13   this show is not that bad but like my [TS]

01:56:15   arse address gets tons of email all the [TS]

01:56:17   time from people are saying my [TS]

01:56:19   computer's broken can you help me fix [TS]

01:56:20   like that's that's paraphrasing but like [TS]

01:56:22   long drawn-out impassioned emails about [TS]

01:56:26   me helping a fix thing appear and almost [TS]

01:56:27   all the time I have no idea how to help [TS]

01:56:30   them I don't even know where to begin [TS]

01:56:31   but the thing is like I'm not their [TS]

01:56:33   personal tech support I'm not gonna you [TS]

01:56:35   know if they ask me a simple question [TS]

01:56:36   like how do you know how to do the blah [TS]

01:56:37   blah and I do I respond I'm not like [TS]

01:56:39   begrudging them some knowledge that I [TS]

01:56:40   have and I'm withholding it but it's [TS]

01:56:42   like I don't know I would have to do the [TS]

01:56:44   same things you have to do I'd be [TS]

01:56:45   googling around figured like I'm not [TS]

01:56:47   going to do your research for you and [TS]

01:56:49   figure out why your computers broken [TS]

01:56:50   take it to the Genius Bar that's what [TS]

01:56:51   those guys get paid for you know so [TS]

01:56:54   emails like that I do not feel any [TS]

01:56:56   obligation to do anything for them it's [TS]

01:56:59   all cost-benefit if I can give you a [TS]

01:57:00   quick answer and say yes respond even if [TS]

01:57:02   it's just like someone saying hey I like [TS]

01:57:04   the show and I respond thanks glad [TS]

01:57:05   you're enjoying it let's I'll do that [TS]

01:57:08   that's no problem is that takes two [TS]

01:57:09   seconds but these you know it's just the [TS]

01:57:12   volume and depth of some of these emails [TS]

01:57:14   can't be addressed and if people really [TS]

01:57:16   do have good points I feel like I should [TS]

01:57:18   address them in a format that my work [TS]

01:57:21   will that will benefit more than just [TS]

01:57:24   the person who's asking the question so [TS]

01:57:25   a blog post or talked about on the show [TS]

01:57:26   or something like that so that's that's [TS]

01:57:30   where I come down so obvious I do have [TS]

01:57:33   some kind of conscience because I don't [TS]

01:57:34   want to leave the people hanging up they [TS]

01:57:35   have some simple questions say