182: I Had to Reboot My Car Today


00:00:00   So, are you recording in a format that is not just call recorder?

00:00:03   Nope.

00:00:04   Start recording in a format that is not just call recorder.

00:00:07   You're just asking me to mess things up.

00:00:09   Oh, because you're afraid that he's gonna pull a casey?

00:00:12   Mm-hmm.

00:00:13   My UPS lasts for at least five minutes.

00:00:15   I'm actually plugged into it, so.

00:00:17   Yes, yes.

00:00:18   Oh, wait, I hear the truck.

00:00:19   Do you hear it?

00:00:20   Do you hear it?

00:00:21   It's driving away.

00:00:22   Let's see.

00:00:23   Maybe it's leaving.

00:00:24   Yeah, I think it's driving away.

00:00:26   All right.

00:00:27   (electronic music)

00:00:29   - So Marco had just asked John if he could start

00:00:31   a crash resilient sound recording program

00:00:34   so that if this truck was doing some sort of electrical work

00:00:37   and suddenly the power went out,

00:00:39   it wouldn't be a big deal because John hopefully

00:00:42   would be using something that isn't call recorder

00:00:44   like I was using when my Mac crapped out.

00:00:47   - Which I can tell you why that matters

00:00:49   if we ever get to talk about the MP3 file format.

00:00:51   - I would love to do that actually, but anyway.

00:00:53   (laughing)

00:00:54   John had said to Marco, "Oh, well it doesn't matter

00:00:56   because my UPS lasts for at least five minutes

00:00:58   and I'm plugged into my UPS.

00:01:00   - That's how I said it.

00:01:01   That was an accurate reproduction.

00:01:03   - That is a completely flawless reproduction.

00:01:05   It was effectively verbatim.

00:01:06   - If only it was recorded.

00:01:08   Oh yes, it was on my end.

00:01:09   On my call recorder recording,

00:01:10   it will successfully make it to disk.

00:01:12   (laughing)

00:01:14   - So anyway, I bring all this up

00:01:15   because not 20 minutes ago,

00:01:18   I was sitting at my iMac,

00:01:20   knowing that I'm going to be going out of town

00:01:23   and thinking to myself,

00:01:25   I don't know what to do because I really want to let this thing run for another week and

00:01:30   see if it reboots itself.

00:01:33   And if it doesn't, I think at that point I will personally be fairly convinced that the

00:01:39   OEM RAM is good and that the OWC RAM was bad.

00:01:44   But I'm probably going to be using Plex at some point.

00:01:47   The Plex server is the iMac.

00:01:49   If this thing turns itself off, I don't have any mechanism to turn it back on.

00:01:56   Should I turn it off and plug it into the battery part of the EPS?

00:01:59   I don't know!

00:02:00   And I decided not to, because I went into System Preferences and confirmed that the

00:02:05   little checkbox that reads "Start up automatically after a power failure."

00:02:10   And so I'm just going on faith that either we won't have a power failure, or if we do,

00:02:15   the Mac will start itself back up.

00:02:17   And somebody in the chat is asking, "You don't shut down your machine at night like the rest

00:02:21   of us?"

00:02:22   No.

00:02:23   This thing is on always.

00:02:24   I turn the screen off.

00:02:25   Yup, I'm the exact same way.

00:02:26   No one should be shutting down their machines at night anymore, ever.

00:02:28   Doesn't make any sense.

00:02:30   To be fair, there's no reason not to, I guess?

00:02:33   No, there is.

00:02:34   There is a pretty big reason.

00:02:35   They do use a good amount of power, but probably the ideal balance between functionality and

00:02:40   power savings is sleep mode.

00:02:43   to actually put the machine to sleep, not the deep hibernation mode, but regular sleep

00:02:47   mode, that is, I think, a great balance for most people.

00:02:50   You know, to put it to sleep, your energy saver settings should do that for you. After

00:02:54   one hour of idle time, go to sleep or whatever. Yeah, but I don't want that happening during

00:02:58   the day, though, when I'm awake. Because at any moment, I might want to do something on

00:03:02   this Mac remotely. And yes, I could do wake on LAN, blah, blah, blah, but...

00:03:06   That's what scheduled sleep is for, then. If you know, schedule to go to sleep at 11

00:03:09   every day and wake up at 5 a.m. and you're fine.

00:03:11   - I didn't know that that was a thing.

00:03:13   - That's a thing.

00:03:14   - Oh yeah, schedule.

00:03:15   - That's how I do my Backblaze backups.

00:03:16   My computer wakes itself up at around 3 a.m.

00:03:19   and does Backblaze and then goes to sleep.

00:03:22   So I don't actually run it continuously all day.

00:03:23   I just do nightly backups.

00:03:25   - Interesting.

00:03:26   - Is that, so I assume the wake up is the Mac feature

00:03:28   and then Backblaze is the thing to say,

00:03:29   put it to sleep when it's done?

00:03:31   - Backblaze has its own independent schedule,

00:03:33   which is either run continuously

00:03:34   or choose when you want me to run.

00:03:35   So I have two independent schedules

00:03:37   I just synchronize good old cron style, say the Mac wakes up at 3, or the Mac wakes up

00:03:42   at 2.55 and Backblaze starts at 3 and yeah.

00:03:46   So every day you're ruining five minutes of power usage of your computer, you're just

00:03:50   wasting that power.

00:03:51   It's not five minutes, it stays awake for like an hour, I let it do all its stuff.

00:03:54   During that time I probably just Time Machine backups and like Power Nap is, I think, predates

00:03:58   or post-dates my Mac, but Power Nap will have things wake up from their sleep and do Time

00:04:03   Machine backups and check mail and stuff, but it doesn't help with Backblaze.

00:04:06   Although technically if it wakes up to do time machine backplays will also run if you

00:04:09   have a set to continue.

00:04:10   So anyway, the point is this scheduling feature is part of OS X.

00:04:12   It's been there for years and years and years.

00:04:14   You should use it.

00:04:15   It's cool.

00:04:16   Yeah, I probably should sleep this thing at night, but I just never do.

00:04:21   And I say that only because nothing is happening on it while I'm sleeping.

00:04:25   I mean, I guess that one of you guys, or maybe like Jason Snell since he's three hours in

00:04:29   the past, might be trying to watch something off of my Plex server that I've shared with

00:04:34   you guys.

00:04:35   In principle, there's no reason I shouldn't just let the thing sleep at like midnight or one o'clock or something like that and wake itself

00:04:40   back up at

00:04:42   Like six or seven in the morning

00:04:44   I mean you can use that as like your own kind of like political statement of you know

00:04:48   You guys should really go to bed like stop watching Top Gear off my flexor

00:04:50   You need to go to sleep

00:04:52   Well, I just thought it was funny because I was having this

00:04:54   This internal debate with myself about whether or not I should move the Mac onto the battery side of the UPS and once

00:05:00   Once I have--

00:05:01   - It should only be a question of when, not whether.

00:05:04   - Well, sure, and that's exactly what I was about to say,

00:05:06   actually, once I have my data point with regard

00:05:09   to the OEM RAM, at that juncture, I will move it over

00:05:13   to the battery side of the UPS, but I don't wanna mess

00:05:15   with anything until then.

00:05:17   - I am still a little concerned that there was that one

00:05:21   seeming GPU-related failure, like a day after you put

00:05:24   the stock RAM back in, that is, that's like the only thing

00:05:28   that's weird about this to me that says, like,

00:05:29   maybe this is a more complex problem.

00:05:32   But the fact that it hasn't happened at all

00:05:34   in what, three weeks or something,

00:05:36   and it was happening about every week?

00:05:38   - That's correct.

00:05:38   - That is a pretty strong switchover there.

00:05:41   So I'm not entirely ready to say it's definitely the RAM,

00:05:45   but I'm probably about ready to say

00:05:49   you should at least get the RAM swapped.

00:05:51   - Yeah, and basically, as I've said in the past,

00:05:55   all I'm doing right now is trying to get a data point

00:05:58   that I can go to OWC or Mac sales, whatever the hell they're called, and say, "Hey, listen,

00:06:03   you know, I was having reboots once a week.

00:06:05   I put the OEM RAM back in.

00:06:06   It ran for blank without a reboot.

00:06:09   I'm pretty confident it's the RAM.

00:06:11   Can you send me new sticks?"

00:06:12   And actually, I don't have the individual's name handy, but somebody had sent me a couple

00:06:17   of tweets over the last couple of weeks saying that they had similar issues with OWC RAM

00:06:21   and ended up returning it, I think.

00:06:24   And I presume at this point I'm well out of the return window because I got this computer

00:06:27   in January. But anyways, they said that they had gotten crucial RAM instead, and thus far

00:06:36   it had been flawless, although to be fair it had only been a few days at that point.

00:06:39   So I think what I'll try to do is I'll try to do a return on the OWC RAM, assuming my

00:06:46   test plays out the way I expect it will, and then if that RAM has similar issues at that

00:06:51   point I'll probably either ask to return it, or I should say an exchange or whatever, and

00:06:56   And then I'll ask to return it if the new hypothetical RAM has the same issue and maybe

00:07:01   just get like, you know, crucial RAM or something like that.

00:07:04   Anyway, we should probably do some follow-up.

00:07:07   Is Jon doing LensRentals wrong?

00:07:09   Yeah, I got an email today that said, this is a reminder from lensrental.com, this is

00:07:14   a reminder that your rental ended yesterday and our system indicates that it has not been

00:07:17   sent back to us.

00:07:19   So I was surprised by his email because first of all, my rental didn't end yesterday.

00:07:23   It ended on Monday, and yesterday was Wednesday, and I sent it back on Monday.

00:07:30   And of course when I sent it back I got a receipt from the FedEx place, and I took a

00:07:32   picture of the receipt with my phone, so as soon as I got this email I could immediately

00:07:36   reply with a picture of my receipt with a tracking number that if they entered into

00:07:39   the FedEx website they would see was on truck for delivery back to their place in Missouri

00:07:44   or whatever they are.

00:07:46   And my question is to Marco, who's done this before, did I do it wrong?

00:07:48   Am I supposed to enter my tracking number after I return it?

00:07:51   Am I supposed to go to the website and click a button that say that I shipped it back?

00:07:54   Did I mess this up somehow?

00:07:56   You didn't mess it up at all.

00:07:57   Basically something messed up that doesn't usually do that.

00:07:59   Anyway, they don't have the lens I want to rent for a moist vacation, so I'm kind of

00:08:03   upset about that too.

00:08:04   Wixia, what lens is it?

00:08:06   It's a lens that I probably don't want to buy because it's $1,000, and I'm trying to

00:08:10   find like a zoom that I wouldn't want to buy because it's like a compromise, but it's a

00:08:14   pretty good compromise.

00:08:15   This is the Sony Vario Tessar T-Star E 16-70mm.

00:08:21   It's a very compact zoom.

00:08:23   It's like if you want to just have one lens on your camera on vacation, it takes decent

00:08:27   pictures at many ranges and has a pretty good zoom range, not really really big zoom, but

00:08:32   you know, pretty wide to pretty close up and folds back to a small size.

00:08:38   This looks like a good lens for that, but then again it's also $1000 and do I really

00:08:42   want to spend $1,000 on a zoom lens that isn't optically that amazing, but it really is very

00:08:48   flexible and it's much smaller than the zoom lens I had. So I'm thinking of renting that

00:08:52   and then just buying some primes with the camera.

00:08:54   Yeah, I mean, honestly, and I was actually very pleased. We got a lot of feedback from

00:08:57   other photographers so far basically saying that they too, and these are some pro photographers,

00:09:04   that they too only own primes. And some of them said I occasionally rent a zoom for like

00:09:09   an event, but for the most part there's been a number of people who are basically saying

00:09:13   that they agree with my plea last episode to please consider only using primes and please

00:09:19   at least get the 50 prime equivalent for your system. There were actually quite a lot of

00:09:23   those. That was very nice to hear. So it does seem like you are looking at that. I also

00:09:28   left a dimension last show. If you're not looking at the Sony system, I would also give

00:09:34   serious consideration to the Fuji XT line or X Pro line.

00:09:39   There was the XT1 that a lot of people loved.

00:09:43   I believe the XT2 is now out or at least it's about to be out.

00:09:47   The whole Fuji X-Trans sensor thing is kind of cool and has some pretty cool advantages.

00:09:52   So I would strongly recommend if you don't want to jump all the way up to Sony's A6300/A7

00:09:59   ranges, consider the Fuji X line because it is, I have not had any direct experience with

00:10:06   it, but it is very well regarded and a lot of people like that a lot at the slightly

00:10:11   lower price point.

00:10:12   People were trying to talk me into micro four thirds, I'm totally aware of all these things.

00:10:17   I think I've read every single review on DPReview for every 2016 camera at this point, so I'm

00:10:22   very aware of the things and you know, I'm pretty settled on the Sony body.

00:10:26   I'm willing to pay the extra money for the extra,

00:10:28   and it's a camera that I use.

00:10:30   It's a known known at this point.

00:10:32   I rented it, I used it, I liked it.

00:10:34   The Fugees are less expensive,

00:10:35   but I don't mind the extra cost for the body,

00:10:37   especially when I'm looking at lenses

00:10:39   that all cost a bazillion dollars.

00:10:41   Anyway, still no purchases made, but I'm thinking about it.

00:10:43   - Also worth pointing out, as one person did

00:10:45   that I also neglected to mention last episode,

00:10:47   the lenses that I buy are the full frame lenses,

00:10:50   the Effie line in Sony's parlance.

00:10:53   John would not need to buy the Effie lens.

00:10:55   he could buy the crop sensor E lenses.

00:10:57   And so like the lens you were just mentioning, John,

00:11:00   that is not available in full frame.

00:11:01   There is no equivalent to that for full frame.

00:11:03   - I know, that's why it's a good lens.

00:11:04   Like it's perfect for my camera.

00:11:06   - That's why it's a versatile lens.

00:11:07   - Well, for my camera, I don't wanna pay

00:11:10   the amount of money I'd have to pay

00:11:12   for a full frame version of the same lens

00:11:14   and I don't want the size that that would bring.

00:11:15   So here is a lens made specifically for my size sensor

00:11:18   that is very compact, that has a good zoom range,

00:11:20   that is optically pretty good.

00:11:21   - Exactly, yeah.

00:11:22   So for your needs, I do recommend for most people

00:11:27   that they, like if they have the budget

00:11:29   for one of these Sony A7 series full frame cameras,

00:11:33   to step up to it if you can

00:11:34   because the quality difference is immense.

00:11:36   However, for what you are using and for your stated needs

00:11:40   for zoom lenses basically, for really far reaching

00:11:44   but also small and somewhat affordable zoom lenses,

00:11:47   I would not recommend full frame.

00:11:48   I think you're making the right move

00:11:50   by not going full frame for that particular reason.

00:11:53   It pains me to say that.

00:11:55   - I would go full frame if the body wasn't so big.

00:11:59   Like I would sacrifice the zoom if the body wasn't so big

00:12:01   and if the body wasn't literally three times the cost.

00:12:03   Like it's not a small price jump.

00:12:05   I'd pay a couple hundred extra, but 3,500 bucks, I'm out.

00:12:08   - And one thing to consider,

00:12:10   if you don't wanna go all the way to the $3,000 A7R II,

00:12:14   the regular A7 II without the R came out

00:12:17   something like six to nine months earlier

00:12:19   And I rented one of those before I decided to wait for

00:12:23   and then buy the A7R II.

00:12:24   The regular A7 II is not that much more money

00:12:28   than the A6300, and it's not quite as advanced

00:12:32   in some of the newer stuff, like some of the video features

00:12:34   and the focus, I think there aren't as many

00:12:37   face detection focus points on it, I think.

00:12:39   - And the burst mode thing, and it's bigger.

00:12:41   Like yeah, I know, I looked at it.

00:12:42   - But the sensor for the A7 II, it is full frame,

00:12:45   and I think, you can get those things for what,

00:12:47   like $14 or $1500, for what they are,

00:12:51   they're an incredible value right now

00:12:52   because it's kind of like last year's model,

00:12:54   that's still for sale basically.

00:12:55   That is a very good buy if you want full frame

00:12:58   but don't wanna spend like three grand on it.

00:13:01   And compared to what was available,

00:13:02   like if you wanted full frame before the Sony A series,

00:13:07   you had to go for like a 5D Mark II, Mark III,

00:13:10   and those were like $3000 cameras.

00:13:12   That used to be the only way to get full frame.

00:13:14   Like that was like the entry point for full frame

00:13:16   was these like $3,000 Canon big SLRs.

00:13:19   To have something like the Sony A7 series

00:13:21   and to have their full frame sensor starting at like

00:13:24   1500 bucks is amazing for quality photography

00:13:27   and for bringing that to people.

00:13:29   It's really quite something.

00:13:30   - Yeah, I'm gonna do the Casey prediction now

00:13:32   and say here's what's actually gonna happen.

00:13:33   I'm gonna get this camera, we're gonna use it

00:13:35   for a few years, the new, whatever replaces your camera

00:13:38   is gonna come out and/or there's going to be

00:13:40   a full frame camera in a similar form factor to the 6300.

00:13:44   And several years after buying this camera

00:13:45   a bunch of lenses, I'm going to sell it all and buy a full frame mirrorless camera, which

00:13:51   may or may not be from Sony. So I know that's going to happen, like it's inevitable, but

00:13:54   I'm willing to spend the three or four years now with this small camera as a transitional

00:13:58   face.

00:13:59   Yeah, that's totally fair. Honestly, I would bet against that future for you, just because

00:14:03   like...

00:14:04   You think I'm too cheap?

00:14:05   No, I think...

00:14:06   Is fair?

00:14:07   Fair. I mean, that might be a factor here, but I've been talking to you about cameras

00:14:11   for like two years now or three years now,

00:14:13   and you've never wavered on the point of like,

00:14:17   I like good cameras,

00:14:18   but I will never buy one that's that nice.

00:14:20   - I know because it's a Money Pit.

00:14:21   I know I'm susceptible to this Money Pit,

00:14:23   and my wife made me enter it,

00:14:25   and as predicted, once you have one and use it a lot,

00:14:28   like you just, it's, I know, I know myself,

00:14:30   I was trying to avoid this particular Money Pit.

00:14:33   Now I'm easing into the Money Pit.

00:14:34   And also I feel like your camera is like,

00:14:37   I feel like your camera is still

00:14:38   a little bit in transition.

00:14:39   I feel like this whole mirrorless revolution is kind of like just getting going here and

00:14:43   I want to, I really believe that you could have a decent full frame camera with all of

00:14:49   your features in a smaller form factor that has even better performance.

00:14:52   I think that is coming, I just have to wait like five to seven years and I'm willing to

00:14:55   wait and the 6300 was really cool and I think I'll use that while I wait.

00:15:01   I'm not sure it's going to get that much smaller honestly because like the A7 line before the

00:15:07   the A7 II, A7 R2, the regular A7 line

00:15:10   was a little bit smaller.

00:15:12   One of the reasons it had to get bigger

00:15:13   was the giant in-body stabilizer thing.

00:15:16   If you look, there's actually a tear down

00:15:18   on iFixit of the A7 R2.

00:15:20   There really is not a lot of room in there

00:15:22   to shrink this thing any further.

00:15:24   And also, if they did shrink it further somehow,

00:15:27   that would probably mean an even smaller battery,

00:15:29   and that's the last thing this thing needs.

00:15:31   - There's plenty of room for more battery.

00:15:33   If they need to talk to Apple about scalped batteries

00:15:35   or something, there's room for more battery in there.

00:15:36   - And you can't change them.

00:15:38   - I'm totally willing to not have the in-body stabilization,

00:15:40   like do the 6300 but with a full frame sensor,

00:15:43   but without the stabilization, like anyway.

00:15:45   I believe in the future of technology

00:15:47   and I believe cameras will get smaller and better

00:15:49   and so I'm willing to wait it out.

00:15:51   And even if it doesn't.

00:15:52   - Wait, if you don't want the stabilization,

00:15:55   hold on, how much is an A7R, the first A7R?

00:15:58   Actually that's contrast only, you don't want that.

00:16:01   - But that's the old camera, slow burst mode

00:16:03   and not as good sensor.

00:16:05   Yeah, you don't want contrast autofocus, never mind.

00:16:08   Yeah, exactly, forget about it.

00:16:09   Anyway, we're getting enough camera talk.

00:16:11   At some point I'll buy something,

00:16:14   then we'll talk about it some more.

00:16:15   Okay.

00:16:16   All right, good talk.

00:16:19   We got a bit of feedback with regard to QNX,

00:16:22   which is that real-time operating system,

00:16:24   and Dan Dodge, I believe is the gentleman's name

00:16:26   that was co-founder and is now working at Apple.

00:16:29   And anonymous, well, two different anonymous sources.

00:16:32   The first told us that the M5 that Marco had, my 335, both the iDrive systems are running

00:16:41   on QNX, and this particular individual also added that there really isn't any requirement

00:16:47   for real-time OS for infotainment.

00:16:49   It just kind of happens to use QNX.

00:16:52   It's kind of like the standard that everybody does.

00:16:54   Yeah, I was about to say, and that's apparently fairly popular.

00:16:57   And furthermore, from what I understand, the CarPlay features that are in a lot of modern

00:17:04   cars, a lot of that support happens at the QNX OS level, even before the iDrive tier,

00:17:14   if you will, or level in the software.

00:17:17   And then a different anonymous source wrote us a really fascinating email, which I'd love

00:17:21   to read the whole thing, but I presume Jon has called out a couple of parts that I'll

00:17:25   read to you.

00:17:26   The biggest performance problem in car UIs is usually

00:17:29   the embedded chipset used.

00:17:30   Our system has to work on 10-plus chipsets by different

00:17:33   manufacturers.

00:17:34   Car companies save every cent on their chipset.

00:17:37   The fact that some cars built today still have less than

00:17:40   128 megs RAM, although RAM costs almost nothing, shows

00:17:43   that perfectly.

00:17:45   As my head of engineering once said to me, they'd rather

00:17:47   spend money on three software engineers for two years trying

00:17:50   to get that system to run as efficiently as possible,

00:17:54   rather than spending a few dollars more on each chip.

00:17:56   And that's because of the economy of scale

00:17:59   that when you're building 11 gazillion cars,

00:18:02   a few dollars per car really starts to add up.

00:18:05   And I just, I mean, it makes sense,

00:18:06   but I didn't really think of it that way

00:18:08   until this anonymous little birdie wrote in and said that.

00:18:11   So I thought that was fascinating.

00:18:12   - It doesn't make sense.

00:18:13   It's the exact definition of Pennywise pound foolish.

00:18:15   Like, because I think like, oh, you're right about the math

00:18:19   and how it multiplies out, but you're wrong about,

00:18:21   or the car companies are wrong about the value

00:18:23   provided by having good software.

00:18:27   The good software is an important differentiator,

00:18:29   an increasingly important differentiator in cars.

00:18:31   And you saving two bucks on that chip

00:18:33   is not worth the value lost to your brand.

00:18:36   Like all the people, like Marco said,

00:18:38   drives this really nice Mercedes,

00:18:41   but it's disgusted by the infotainment system.

00:18:42   You spent all this money on this car.

00:18:45   So many parts in the Mercedes are 50 cents more expensive

00:18:48   than they are in other cars.

00:18:49   One cent more expensive, $1 more expensive

00:18:51   from like the fuel lines to the filters,

00:18:53   to the mufflers, to every single part of that car

00:18:55   is just a little tiny bit more expensive than Honda.

00:18:57   That's what makes it a Mercedes.

00:18:59   The software you can't treat like,

00:19:00   "Oh, well, we're only gonna put 120 megs of RAM

00:19:02   because doubling the RAM would cost three extra cents

00:19:05   and instead we're gonna spend, you know,

00:19:07   have three engineers work on this

00:19:09   and to jam the software into this tiny little chipset

00:19:11   with a small amount of RAM."

00:19:12   Because hey, it's cheaper, do the math.

00:19:14   It is cheaper, but you're making your product worse

00:19:16   in a way that is way out of proportion

00:19:18   to the money that you save.

00:19:19   If you just add up like, okay, two bucks,

00:19:21   two extra bucks per car times the number of cars we sell,

00:19:23   you're adding so much more value to your brand as Mercedes,

00:19:26   as a car that people wanna buy by making decent software.

00:19:29   And you make decent software

00:19:30   by not making your software engineers

00:19:32   have to jam their software into a bunch of weird,

00:19:35   underpowered chipsets that are RAM constrained.

00:19:38   It's a terrible idea.

00:19:39   You know, this email ends to say,

00:19:41   the good part is all these problems

00:19:43   are slowly getting worked on.

00:19:44   The companies creating these infotainment systems

00:19:47   are slowly getting better at understanding

00:19:48   the technology hurdles.

00:19:49   And some car companies that really care

00:19:51   are taking this away from third party vendors,

00:19:53   basically taking it in house, kind of like Tesla does.

00:19:56   So they're getting better about this,

00:19:57   but the mindset is like totally wrong.

00:19:59   And we're totally not blaming the people

00:20:01   who make these infotainment systems,

00:20:02   because as this person said, they have to write this thing,

00:20:05   it has to work on these terrible washing machine chipsets

00:20:08   that are in these cars because they're cheap.

00:20:10   And of course they're terrible.

00:20:11   And by the way, the infotainment system on QNX thing,

00:20:13   that was in response to the last episode

00:20:16   where we were talking about QNX, and it's like,

00:20:17   well, you don't need a real-time operating system

00:20:19   for the infotainment, you need it for like the self-driving car, the automatic stopping

00:20:24   system that prevents you from hitting pedestrians, and other things that are critical that need

00:20:27   to be in real time.

00:20:28   But QNX is also a reliable, popular embedded operating system, so I guess they just use

00:20:32   it for the infotainment systems too.

00:20:33   Maybe they're not even using the real-time features.

00:20:35   This was my question.

00:20:37   What is it about the infotainment system that needs real time?

00:20:40   And the answer is nothing.

00:20:41   It's just a good embedded operating system that automotive engineers are familiar with

00:20:45   that runs on those washing machine chips

00:20:48   with a tiny amount of RAM, so there you have it.

00:20:50   - I should also point out that I had to reboot my car today.

00:20:53   - How do you do that?

00:20:54   - It happens, you hold down a key combination, right?

00:20:57   - Yes, and I should point out too,

00:20:59   this is the third time I've had to do this.

00:21:01   And since owning the car for three months.

00:21:04   - You should really look at the RAM, Marco,

00:21:06   I hear that's troubling these days.

00:21:08   - Yeah, so.

00:21:09   - So what was the symptom?

00:21:10   - So the first time I had to reboot it,

00:21:13   so there's two screens, there's the screen

00:21:14   where the speedometer is right in front of you

00:21:16   on the little dash console thing,

00:21:18   and there's the giant touchscreen in the middle.

00:21:21   And those are actually running two different computers,

00:21:23   or the one in front of the driver is kind of

00:21:26   like a sub-interface that remotely controls

00:21:29   the main one in the center, and the main one

00:21:31   in the center also controls all the HVAC stuff,

00:21:35   all the media stuff, and a bunch of other things

00:21:39   that are not driving, but are a lot

00:21:40   of other accessory features of the interior.

00:21:43   So I went out one afternoon and the driver facing one said,

00:21:48   there's a problem with your center console, call service.

00:21:51   I hate talking on the phone,

00:21:52   so rather than calling customer service,

00:21:54   I just searched the web for this error message

00:21:56   and found some forums, posts and everything saying,

00:21:59   oh, just reboot it.

00:22:00   And then it took me another 10 minutes

00:22:01   to figure out how the heck do you reboot it.

00:22:03   And the way you reboot it is,

00:22:05   the steering wheel has a little jog wheel on each side

00:22:07   like by the buttons for the controls.

00:22:10   And if you click in both of those jog wheels

00:22:12   and hold them in for like five seconds,

00:22:15   the center screen reboots.

00:22:16   (laughing)

00:22:17   - Interesting.

00:22:18   - So I did that, it fixed the problem,

00:22:20   you know, it came back up,

00:22:21   it takes like 20, 30 seconds to reboot,

00:22:23   came back up, it was fine.

00:22:25   And then I did it willingly about a month ago

00:22:29   because the navigation, when it highlights

00:22:33   the road you're supposed to drive on with a blue overlay,

00:22:37   sometimes if you zoom or pan

00:22:40   or otherwise cause the map to redraw,

00:22:43   sometimes the blue overlay stops drawing reliably.

00:22:46   It'll either omit sections or it'll just not draw at all.

00:22:50   And it looks like it's maybe like running

00:22:51   out of texture memory or something,

00:22:53   and so like the texture is just not being painted right

00:22:55   or something.

00:22:56   I don't know how these systems work in enough detail

00:22:57   to say for sure, but it looks like some kind of just like

00:23:00   memory leak bug that slowly this just stops

00:23:02   working very well.

00:23:03   And actually when I had the service loaner,

00:23:05   when I got my windshield replaced,

00:23:06   that car had that same symptom where it just wasn't

00:23:08   drawing the overlay right.

00:23:09   so I know it's not just my car.

00:23:11   I used to know how to fix it in the service owner,

00:23:12   but now I do, you just reboot it, it's immediately better.

00:23:15   And then today, the center console just stopped responding

00:23:18   and the audio froze and it was playing over Bluetooth

00:23:21   so it wasn't like a cable issue.

00:23:22   And yeah, so I was driving at the time,

00:23:24   so I'm like, well, I hope this doesn't do anything weird.

00:23:26   So I was stopped at traffic lights, I'm like,

00:23:28   all right, good time, hold 'em in, reboot.

00:23:31   It was fine, came back up again, totally fine.

00:23:35   I will say overall, now that we're on the subject

00:23:37   of these infotainment systems, it kind of annoys me

00:23:40   that Tesla is working on all these automatic driving things

00:23:44   and pouring so much of their resources into that

00:23:47   while their in-car navigation and media system

00:23:51   still could use a lot of improvement.

00:23:53   It's still pretty rudimentary.

00:23:54   It is nice in that it has the big touch screen

00:23:57   and everything, but the media interface is extremely basic.

00:24:02   It supports very little of Bluetooth audio capabilities,

00:24:06   control capabilities, doesn't support like artwork

00:24:08   and stuff like that, doesn't support fast forward

00:24:10   and rewind, there's no of course browsing or anything,

00:24:13   no iPod interface, no CarPlay, so they're missing a lot

00:24:17   there and then the navigation is also very rudimentary.

00:24:21   You can't set waypoints, you can't like say,

00:24:24   all right, well drive here to here but use this highway

00:24:26   or drive here to here but stop here in the middle.

00:24:28   You can't do that, you have to just make separate trips.

00:24:31   So stuff like that and also the directions,

00:24:33   Like the map images come from Google,

00:24:37   but I think Tesla has its own street data,

00:24:41   and Tesla also provides its own navigation,

00:24:44   despite wherever the data's coming from.

00:24:46   Like it's not using Google's navigation,

00:24:47   Tesla has its own navigation.

00:24:49   And Tesla's time estimates are never right with traffic,

00:24:52   and it is not very good navigating

00:24:55   around traffic-heavy cities.

00:24:56   Like I've made the mistake of following it

00:24:57   a couple times for Manhattan,

00:24:59   and it was a disaster every time.

00:25:02   And it was not that driving in Manhattan

00:25:04   had to be a disaster, it just made really bad choices

00:25:06   and I followed them.

00:25:07   And then I would look over and like,

00:25:09   oh, if I would have just stayed on the highway

00:25:11   one more exit, I would have bypassed all of this.

00:25:13   That took me 40 minutes.

00:25:15   So overall Tesla's head unit is,

00:25:19   or like the Navigate system is, it's good.

00:25:22   I love having the big screen.

00:25:24   However, I wish they'd put more resources into that

00:25:27   because it seems like they haven't really touched that

00:25:30   in a long time.

00:25:31   and for them to be putting all this amazing effort

00:25:34   into cool AI features and stuff, that's nice.

00:25:37   And if you look at the big picture,

00:25:39   if these things end up saving lives, that's awesome,

00:25:42   and that is more important.

00:25:43   But at the same time, as a customer of their car,

00:25:46   I do wish the system was better.

00:25:48   And it is kind of weird that such a tech-forward company

00:25:52   has kind of dropped the ball on the basic features

00:25:56   of their technology in the car facing the user.

00:26:00   Also, the touch screen is just really sluggish.

00:26:03   Navigating the map is very, very low frame rate.

00:26:07   It's kind of like navigating Google Maps on a Pentium 3.

00:26:12   It's not like using an iPad.

00:26:14   It's not at all like that.

00:26:16   It's very sluggish and there's all this latency

00:26:18   in the interface and you do kind of expect more

00:26:22   from a car of that caliber.

00:26:23   - So features aside, getting back to the rebooting

00:26:27   your car and stuff, you kind of, well, I don't know, it's hard to separate it from the features

00:26:32   when talking about this, but I think a lot of regular people have this feeling, and it's

00:26:37   easy for us to slip into it too, it's like, when things were simpler and didn't have as

00:26:42   much computer stuff, they were more reliable because there were fewer things that can go

00:26:46   wrong, and as soon as they started adding software to their stuff, like when we talked

00:26:50   about this with smart TV, that old smart TV post I had from CES, so worst products through

00:26:54   software. They had all this software, and it just makes the thing less reliable and

00:26:58   more annoying to use. And you're speaking of it in terms of the basics, like, they've

00:27:04   added a bunch of software to the car, which is good, and they're better at it than a lot

00:27:07   of other people, but on the other hand, it's also less reliable. In my terrible, terrible

00:27:11   infotainment, if you want to even call it that system on my Honda Accord, I've never

00:27:15   had to reboot it, but it's just terrible all the time. So that's kind of like in the in-between

00:27:19   phase where everything is not as simple as it used to be, like on my wife's Accord that

00:27:23   has no infotainment screen at all.

00:27:25   That was, you know, it's simple, fixed function,

00:27:27   everything implemented in hardware

00:27:29   with some very rudimentary firmware, you would even call it.

00:27:33   No sort of general purpose touch screen menu system software,

00:27:37   like nothing like that.

00:27:38   There's a bunch of CPU's in there,

00:27:40   there's a bunch of memory,

00:27:40   but they're all basically like little embedded systems.

00:27:43   That's like as far as the car industry can go

00:27:46   along the lines of like everything is super reliable.

00:27:48   And as we all know as programmers,

00:27:50   but it's sometimes easy to forget as consumers,

00:27:53   As soon as you make real live software like a GUI, you can't make it the same way as an

00:27:58   embedded system.

00:27:59   You have to have a GUI framework and an API and regular application development for human

00:28:04   interaction as opposed to this is custom software for this little chip that controls the radio.

00:28:13   The more you start making a platform, the more you start making what we know is like

00:28:16   a PC style platform where you build applications on top of it where you're farther away from

00:28:21   that ideal embedded system that just has like a ROM or something and you just get all the

00:28:25   bugs out of it and you get it right.

00:28:27   If there are bugs, there are sort of known bugs.

00:28:28   Once you start having real live software, there's nothing so far that we as humans have

00:28:34   been able to figure out how to do to make that software as reliable as the thing without

00:28:41   software without sacrificing huge things like just, okay, we'll do space probe reliability

00:28:49   tricks, but it will take us years and years to write a very small amount of code and the

00:28:53   limitations are so onerous that no one would ever want to do anything that way.

00:28:56   Unless you're writing a space probe, in which case you have no choice, right?

00:28:59   But cars are not space probes, the cost-benefit ratios are different.

00:29:03   So bottom line is, once you add software to cars, you have to reboot your car.

00:29:06   Like essentially that is inevitable.

00:29:08   We do want software to be added to our cars, because we think it's a better user interface

00:29:12   and they can do a lot of stuff, but you will have to reboot your car.

00:29:14   Like there's no, like, if only you had tried harder Tesla, then Marco wouldn't have had

00:29:18   to reboot his car. There's a sliding scale of quality, but a car with software is never

00:29:25   going to be as reliable as the car with the hardware radio, because the hardware radio

00:29:31   either works or it doesn't, and when it stops working, you throw it out and you put in a

00:29:33   new hardware radio, and for the life of that hardware radio, it doesn't change, there's

00:29:36   no firmware updates, there's no software being sent to it, and it's doing so much less, it's

00:29:39   doing so, so much less than Marco's thing. It's like pulling images and doing GPS and

00:29:44   navigating and talking to you and putting images up on the screen. It's doing so much

00:29:48   much less. Like it's not as if, you know, it's this magic sauce. When you do more stuff,

00:29:52   it's more complicated. Software has bugs. The more software you have, the more bugs

00:29:55   you have. And until we get like a sort of emergent AI to write and fix our software

00:30:00   for us, that seems to be the way of things probably for most of our life. So we should

00:30:05   just kind of get used to this. Now, the features discussion, I think, is a separate issue,

00:30:08   which Mark was saying, like, you know, you're going to have bugs, you have a lot of software,

00:30:12   spend more of your time here instead of less of your time there. And we say the same thing

00:30:15   about Apple, especially on the Mac.

00:30:17   If you're not, don't try to look for the next big feature, Ed.

00:30:19   Why don't you just make all the features that are already there much more reliable?

00:30:22   And that definitely is possible, and it is something Tesla should be doing.

00:30:25   But whenever I feel high and mighty about the fact that I never had to reboot my cord,

00:30:29   I realize it's only because it doesn't really have as much software and features as

00:30:32   Marco's car does.

00:30:33   And as soon as it does, it's going to be just as bad.

00:30:37   And we as programmers understand why that happens, because software without bugs doesn't

00:30:42   exist.

00:30:44   More software, more bugs.

00:30:46   - I see what you did there.

00:30:47   - Wow.

00:30:48   - Is that depressing?

00:30:49   I don't know, but like, you know,

00:30:51   that's the way of things.

00:30:52   I feel like we should understand that better than anyone

00:30:54   because we do it for a living

00:30:56   and we know what it's really like.

00:30:58   - We are sponsored tonight by Hover.

00:31:02   Go to hover.com, which is the best way

00:31:03   to buy and manage domain names

00:31:05   and use coupon code RIPTivo for 10% off your first purchase.

00:31:10   Now, finding a perfect domain name

00:31:13   is ridiculously easy with Hover.

00:31:14   So when you have a great idea for your blog or store

00:31:17   or a new startup you wanna do,

00:31:18   start with a great domain name

00:31:19   and then you can just move on.

00:31:21   And the great thing is with Hover,

00:31:22   it's incredibly easy to set up your domain name

00:31:24   with the most popular website host and website builders.

00:31:27   They have a new feature called Hover Connect.

00:31:30   Lets you set up your domain automatically

00:31:31   with tons of popular providers in just a few clicks.

00:31:35   So you don't have to like dig through help articles

00:31:37   trying to figure out all the DNS settings and everything

00:31:39   trying to get your domain working with your web host.

00:31:41   Hover has so many of them built in,

00:31:43   chances are you don't have to do very much at all,

00:31:44   only a few clicks.

00:31:46   And if you need any help,

00:31:47   Hover has amazing customer support.

00:31:49   You can of course email them and everything else.

00:31:52   You can also call them on the phone.

00:31:54   If you call them on the phone during business hours,

00:31:56   an actual human being picks up the phone,

00:31:59   no hold, no wait, and no transfers.

00:32:02   That human being who picks up the phone

00:32:05   is able to help you with whatever questions you have.

00:32:07   No annoying phone trees or voice menus or touch tone menus

00:32:11   or transferring you between a whole bunch of people

00:32:12   trying to find the right person.

00:32:14   No hold, no wait, no transfer.

00:32:16   Amazing phone support if you need it at Hover.

00:32:19   So check it out today.

00:32:20   Go to hover.com when you wanna find a domain name

00:32:24   for your idea.

00:32:25   This week, use promo code RIPTivo at checkout.

00:32:28   That's RIPTivo to save 10% off your first purchase.

00:32:32   Thank you very much to Hover for sponsoring our show.

00:32:38   Jamie has written in and they asked, "I'm surprised that the clocks in different computer

00:32:43   recording devices can't keep frame accurate time. I say frame since I come from the film

00:32:47   world, but you know, accurate to a 24th or 30th of a second. This has been solved in

00:32:52   sound recorders and cameras since the 50s and 60s, using a quartz crystal with the inaccuracy

00:32:56   of a few parts per million. Computers can't provide the accuracy of a quartz watch. It's

00:32:59   so nuts." Marco, would you like to comment?

00:33:02   >> Marco: Yeah, so this is in response to my discussion about audio drift being a problem

00:33:06   and long podcast recordings where basically when your sound card is sampling your microphone

00:33:13   at 44,100 times per second, then over the course of like a two hour recording, what

00:33:20   your computer thinks is the correct time and the number of samples it has recorded might

00:33:24   be off from your co-host's computer by like half a second to one second. And it's just

00:33:31   due to imprecision between the computers.

00:33:33   Like, my sound chip might just be, you know,

00:33:36   .0001% less accurate than yours,

00:33:39   but when you're doing 44,100 samples per second

00:33:42   over two hours, that could start to add up

00:33:44   to something noticeable, 'cause it only takes

00:33:46   like a half second for the two tracks to be off

00:33:49   before you can really notice it.

00:33:50   My reply to Jamie's feedback here is,

00:33:53   I don't know why they can't do it.

00:33:54   I mean, if it's possible to do it somewhere else,

00:33:56   I don't know why, I'm sure there's a good reason,

00:33:58   'cause this has been a problem for a long time.

00:34:00   One of the ways that this problem gets avoided

00:34:02   in professional equipment,

00:34:03   which Jamie's probably familiar with,

00:34:05   if you look at pro audio gear,

00:34:07   like really pro audio gear,

00:34:09   there is usually a clock input and output port on the back,

00:34:14   and you can use separate devices for clock generation.

00:34:17   So what they do is they don't rely on incredible precision

00:34:21   to make sure that this device and this device

00:34:23   have exactly the same precision on their clock chips.

00:34:26   No, they just outsource the problem.

00:34:28   They declare like one source of the clock rate

00:34:32   to be authoritative, and they just hardwire

00:34:35   all to each other and say, all right,

00:34:37   this device is providing the clock for me,

00:34:39   so my internal circuitry's not gonna do it.

00:34:40   This is the world I know very little about.

00:34:42   I just know that this exists, so if I'm getting

00:34:44   those details wrong, I apologize, but that's the gist of it.

00:34:47   That is how that world deals with it.

00:34:48   It's not by getting incredibly processed components.

00:34:51   It's simply by outsourcing the problem to one device

00:34:54   and saying, okay, you're the master of the clock.

00:34:57   I put this feedback in there because I had the same question when you were describing

00:34:59   it and I was thinking along the lines of, look, the clocks have to be pretty darn accurate

00:35:04   from one computer to the next.

00:35:06   Could it be instead that, because again we're not using real-time operating systems here

00:35:10   to record our audio, that it's easy to miss a frame here and there because some other

00:35:16   thing happens and it's not a big deal, you'll never notice it audio-wise, but if you add

00:35:20   that up, like in other words it's basically a software problem and not a hardware problem,

00:35:24   the quartz crystals in our computers are all close enough to each other that it wouldn't

00:35:29   matter but our computers aren't hardware audio recording devices, they're general purpose

00:35:33   computers and in the end it's a piece of software that's pulling samples off a bus and so many

00:35:37   things in the middle of it, missing a few things here and there. But I guess that wouldn't explain

00:35:42   systemic drift because if it was really a random error and it was equal on all our computers maybe

00:35:46   it would even out anyway. I don't know the answer either. Your theory sounds reasonable but I have

00:35:51   have the same questions about like look quartz crystals should be accurate and

00:35:55   matching but then again I didn't know about this whole thing of having a clock

00:35:57   being fed into multiple audio components so I don't know what to think so if you

00:36:01   are an audio person who does audio on computers and you know the definitive

00:36:04   answer please write in and tell us yeah that'd be great also for whatever it's

00:36:07   worth I've always noticed that laptops have way more drift than desktops Mac

00:36:14   pros have way less drift than IMAX like so it does seem to possibly be related

00:36:20   just like kind of the quality and the conditions of the components.

00:36:25   I wouldn't say like the components like a laptop may just be under more stress like

00:36:29   it's more likely to be thermal throttling up and down and doing other stuff like again I would

00:36:33   imagine that the clock crystals inside the related chips and all those things are of equal quality

00:36:38   but it's not it's a we are not solid state audio recording devices like a little handheld thing to

00:36:44   do it's a whole computer in there so so much can go wrong between the samples being pulled off this

00:36:49   this really complicated bus and going out to a file.

00:36:52   So anyway, please write in and tell us.

00:36:54   But we all see the results.

00:36:56   If you want to try to experiment yourself, feel free.

00:36:58   Do a double-ender podcast with two friends across the country

00:37:01   on a laptop and try to put the audio together

00:37:02   and you'll realize why you need Marco's thing.

00:37:05   - All right, let's see here.

00:37:07   So one of you put in here, and I'm quoting,

00:37:11   "The ARM Mac feedback that won't die."

00:37:14   - It won't.

00:37:15   (laughing)

00:37:16   People keep suggesting it.

00:37:17   It's been like, what, three months? Six months? This literally will not die.

00:37:21   And I feel like now, okay, fine, we will address it.

00:37:24   Quote, "It's obvious what's going on with the Mac lineup," colon, "Total ARM reboot."

00:37:29   This is from a tweet. This is a concise form of what everyone is telling us every time we complain about,

00:37:34   "Where are the new Macs? Why aren't they updating the Mac Pro? What's taking so long with the Mac Pro updates?

00:37:38   Everything says 'Don't buy' in the Mac buying guide thing on the MacRumors site.

00:37:42   Doesn't seem like they care about the Mac anymore." Everyone's like, "You dummies!

00:37:45   It's because every single Mac is being converted to ARM

00:37:47   and they're just taking a long time.

00:37:48   Well, I don't understand why you don't realize that.

00:37:50   Why don't you reply to my tweets,

00:37:51   "All the Macs are going ARM."

00:37:53   (laughing)

00:37:54   Maybe, maybe all the Macs are going ARM.

00:37:57   We've talked about that many, many times in the past.

00:38:00   A couple things I have to say about that

00:38:03   and the reason why we don't bring that up.

00:38:04   First of all, we've talked about ARM Macs forever

00:38:06   on past shows and all the different trade-offs inherent

00:38:08   and then we're not gonna rehash all that, right?

00:38:11   But first, if all the Macs are going ARM,

00:38:14   still kinda doesn't excuse the tremendous delay.

00:38:17   Like, look at the Intel transition.

00:38:20   They didn't stop selling PowerPC Macs

00:38:22   and not update them for three years

00:38:23   before the transition to x86, right?

00:38:26   And they knew that was coming for a long time.

00:38:27   You sell the old computers right along

00:38:30   until you tell the world,

00:38:31   "Hey world, we're changing to a new architecture."

00:38:34   - Well, in all fairness though, that's a bad example

00:38:36   because back then, the Mac was Apple's main product.

00:38:40   Like, they cared a lot more about it.

00:38:42   prioritize it a lot more engineering wise, these days it doesn't seem like it is that

00:38:46   high of a priority.

00:38:48   But if it's not a high priority, it's not a high priority whether it's changing or not.

00:38:51   And that gets back to last week's argument of like maybe it's not a high enough priority

00:38:54   to care to switch to ARM because then you'd have to develop ARM chips suitable for all

00:38:57   your Mac line and that costs a lot of money.

00:38:58   But anyway, the second thing is, and the conventional wisdom about the ARM transition is, if Apple

00:39:05   is going to convert to ARM, they have to essentially pre-announce it because developers need to

00:39:09   to get their stuff together,

00:39:10   the same way they pre-announced the x86 transition.

00:39:12   They didn't announce it and say,

00:39:13   "And you can buy an x86 Mac today."

00:39:15   Nope, they had to announce it to developers first,

00:39:17   and then developers could get like that test hardware,

00:39:19   it was like a Pentium 4 in a cheese grater case,

00:39:22   and you could make sure you recompile your apps

00:39:25   and update all the tools and blah, blah, blah.

00:39:27   So could be, could be that Apple

00:39:29   is converting their whole line to ARM,

00:39:32   but that is not the obvious explanation of what's going on,

00:39:37   And it's not even the number one most likely explanation.

00:39:39   It's probably like the number two or three

00:39:41   possible explanation.

00:39:43   And I would be very surprised if Apple

00:39:45   was converting the entire Mac-lined ARM

00:39:48   and didn't tell the world and developers

00:39:51   before you could buy the hardware.

00:39:52   So I don't know, anyone, you have anything else to add

00:39:55   about this eternal feedback that tells us we're adults,

00:39:59   we're not realizing it's a complete ARM reboot?

00:40:01   - Like it's a nice idea if you don't think about it

00:40:03   for too long or if you're not that familiar

00:40:05   with what would be involved or like the markets here.

00:40:08   I just don't think it's realistic.

00:40:09   I don't think there's, not to say that Apple couldn't

00:40:13   release a whole line of ARM Macs, but that--

00:40:15   - They totally could.

00:40:16   - Yeah, as we've talked about, and we're all

00:40:19   under the impression that they probably have

00:40:21   ARM OS X running in the labs and have for some time,

00:40:23   they probably just maintain it as like a parallel

00:40:26   just in case branch, but there really is just not

00:40:29   a lot of motivation for them to do that right now.

00:40:31   As we discussed, I'll summarize it briefly,

00:40:34   basically it would take a lot of work,

00:40:36   would have a lot of downsides,

00:40:38   and Intel just isn't bad enough yet.

00:40:40   - Yeah, and again, what I get carrying back to is like,

00:40:44   ARM Macs, yes, no, whatever,

00:40:46   but the delay in updating all the Macs is unrelated,

00:40:50   because there's no reason why you would say,

00:40:52   because we're doing the ARM thing,

00:40:54   therefore we're gonna delay.

00:40:55   It's not as if they would say,

00:40:56   we don't have time to update those old cruddy Macs,

00:40:58   we're gonna take every person

00:41:00   who was working on Mac hardware before

00:41:02   and put them all on the ARM hardware,

00:41:04   that, you know, it doesn't make any sense.

00:41:07   Also because there have been massive delays in the past,

00:41:09   like the Cheesegrader Mac Pro not being updated forever,

00:41:11   and that wasn't explained by an ARM reboot either.

00:41:14   It's just plain old, like,

00:41:15   there's so many more explicable reasons about Skylight

00:41:17   having bugs and taking a long time to come out

00:41:19   and being delayed and them skipping generations,

00:41:21   like those are the actual reasons why there's a delay.

00:41:23   Even if they come out with a complete ARM reboot in a month,

00:41:26   I will still say this theory was wrong.

00:41:28   Why?

00:41:29   Because all those delays were not related

00:41:30   to the ARM reboot, they're related to all the things

00:41:31   we've talked about so many times.

00:41:32   are not—it's not speculation that they skipped chip generations. They did. They skipped

00:41:36   them. And then it's not speculation that Skylight was delayed and had rollout problems.

00:41:39   That's a real thing that happened. That alone is sufficient to explain the delay.

00:41:44   And what we're complaining about on the past shows is, "Hey, Apple, don't skip generations,

00:41:48   because if you do, any bump in Intel's plan makes your computers embarrassingly late."

00:41:53   Bingo. And by the way, update your GPUs.

00:41:55   [laughter]

00:41:56   (laughing)

00:41:58   - Well, and the other thing I just wanted to throw in there

00:42:01   really quickly is that it's been, for me personally,

00:42:05   less of an issue since my new job

00:42:07   and since I'm doing iOS development.

00:42:09   But in my last job, and a job or two before that as well,

00:42:14   all of the developers, generally speaking, used Macs.

00:42:18   But they all lived in VMware Fusion or Parallels

00:42:22   or VirtualBox if they really hated their lives.

00:42:26   And virtualizing against the same chipset is fairly easy.

00:42:31   You know, to have OS X running on x86

00:42:34   and then virtualize Windows, which is also running on x86,

00:42:37   that's fairly straightforward and easy.

00:42:39   If this was an ARM Mac trying to virtualize x86,

00:42:43   that would have just slowed everything down tremendously

00:42:45   in all likelihood, and would really be a potentially

00:42:50   very bad thing for business users that need to have VMs.

00:42:54   even business users that need VMs not for their day-to-day,

00:42:57   but for their one old legacy app that only runs on IE6.

00:43:01   And so they need to boot into XP and a VM

00:43:03   to use that one app.

00:43:05   Like that happened a lot in past jobs.

00:43:08   And maybe if it's just that one app in IE6,

00:43:11   you can deal with it being slow.

00:43:12   But I mean, up until this job, I lived in VMware Fusion

00:43:15   and I was developing in VMware Fusion.

00:43:18   And it would probably be a lot worse

00:43:21   if the Mac in which I was using wasn't Intel.

00:43:24   And to the point that I would probably have to use

00:43:26   some crappy Dell or Lenovo or something like that.

00:43:29   And that would be just sad.

00:43:30   - Yeah, Microsoft did take a run at this ARM transition

00:43:33   with whatever the Windows for ARM thing

00:43:35   for their Surface thing, and that's not going so well.

00:43:38   And they did it badly, they didn't commit.

00:43:40   They're just like, well, we're also gonna have

00:43:41   an ARM version of Windows, that's cool, right guys?

00:43:43   And the market was like, not that cool

00:43:47   'cause we need to run our x86 software.

00:43:48   Yeah, the only way to do it is the way Apple does it,

00:43:50   which is like, look, we're changing everything

00:43:51   from 68K to PowerPC, get on board,

00:43:54   and if they do an ARM transition on Mac,

00:43:55   I feel like they're gonna do the same thing.

00:43:57   They're not gonna let the two computers live on,

00:43:58   which will mean exactly like Casey said,

00:44:00   if you've gotta run that x86 stuff in virtualization,

00:44:02   you're gonna be super sad,

00:44:03   and it's not gonna be feasible.

00:44:06   It's gonna be like the bad old days

00:44:08   when I ran virtual PC to run x86 software on PowerPC,

00:44:11   and it was so slow.

00:44:13   Oh, just the worst.

00:44:14   - Do you remember, I think we talked about this here

00:44:17   at some point, but they had the daughter cards

00:44:20   for old, old, old Macs.

00:44:21   I remember hearing about this when I was a kid.

00:44:23   - Yeah, but they had 486s on them and stuff.

00:44:25   - Yeah, it was basically just a whole PC on a card.

00:44:27   - Yep, that's what it was. - Right, right.

00:44:29   And this was just so that you could have

00:44:32   a somewhat livable virtualization experience on ancient Macs.

00:44:36   All right, one last piece of follow-up,

00:44:37   and then we are finally done.

00:44:39   - I can't believe we're still in follow-up.

00:44:40   - I know, especially since we recorded, what,

00:44:42   two nights ago, three nights ago?

00:44:43   How is this possible?

00:44:44   - These are topics, come on, these are really just topics.

00:44:46   No, these are all—I can justify every single one of these for being follow-up, and so can

00:44:51   you, because you know what they're about. Apple is follow-up. We talked about it last

00:44:54   episode.

00:44:55   Oh, stop it. I have issues with long follow-up, but it just so happens that we have long follow-up.

00:44:59   Gwyneth Paltrow and Will.I.Am, which we sort of kind of knew, and Gary Vaynerchuk—I hope

00:45:04   I pronounced that right—

00:45:05   You did.

00:45:06   —are going to be the advisors for Planet of the Apps. That was just announced—was

00:45:09   that today or yesterday? Sometime recently, as we record this.

00:45:13   as a—Gwenyth Paltrow is going to be a mentor to the contestants, so I guess she's there

00:45:17   for just moral support and motivates them. I'm not quite sure what her credentials

00:45:22   are to help people develop applications, but you go, guys. Anyway, and she's famous.

00:45:27   And Vaynerchuk is a VC, so I guess he's on the VC side of the Shark Tank fence. Will

00:45:34   I Am is a famous person who likes Apple, who has worked with Apple a lot, and he's an

00:45:38   entertainment industry person, and I guess that's relevant to building apps if you're

00:45:41   affiliated entertainment, and they will serve as advisors,

00:45:45   which I'm not sure is different than a mentor.

00:45:47   Anyway, if you're wondering who they're gonna get

00:45:48   to be on the show, Vaynerchuk, I could've,

00:45:50   like that's right up the middle of what I would expect

00:45:52   in VCs who want to promote themselves and their VC-ness.

00:45:55   - Well, Vaynerchuk is not quite really a VC.

00:45:59   He does his own stuff, he's like a business consultant.

00:46:02   He came up through Wine Library TV,

00:46:05   and then came up and developed this promotional company

00:46:08   with a bunch of his own media projects,

00:46:10   He's basically a media personality and a business speaker

00:46:14   and business consultant and everything.

00:46:15   - Did he do Corked with Dan Benjamin

00:46:17   or was that a different guy?

00:46:18   - No, that's Dan Siederholm, I believe.

00:46:20   But Gary Vaynerchuk, he got famous by having this

00:46:25   really quite amazing very early video series,

00:46:28   I don't even know if it was on YouTube,

00:46:30   I think it was on YouTube but I'm not positive,

00:46:32   called Wine Library TV.

00:46:33   His family owns this big wine retailer in New Jersey

00:46:38   and every day he would just go and do this five minute video

00:46:43   featuring a couple of wine reviews

00:46:45   from a really New Jersey middle class approach to wine.

00:46:50   So it was very low BS, reviewing low cost stuff,

00:46:55   and he was very good at describing how it tasted,

00:46:58   like oh this tastes like grass.

00:46:59   He was surprisingly charismatic and excellent

00:47:03   at this seemingly boring topic of wine reviews.

00:47:07   Anyway, so then his whole career kind of ballooned after that

00:47:10   of being this kind of outrageous, loud guy

00:47:13   who promotes stuff and talks to businesses

00:47:16   about how to do stuff.

00:47:17   But actually, I would say of this list of people,

00:47:20   he is probably by far the most qualified to advise people

00:47:24   on how to do their app stuff.

00:47:27   Although, I admit I have not followed his work

00:47:29   very closely recently, but from what he has done

00:47:33   in the past, he's definitely more related

00:47:36   and is better at knowing how to promote stuff.

00:47:40   - Real time follow up, Gary Vaynerchuk did buy corked.

00:47:43   He bought it from the two that,

00:47:45   Dan Zenerholm and Dan Benjamin.

00:47:47   Yeah, and by the way, I found this out by Googling

00:47:49   and selling the site winecast.net.

00:47:53   Brief aside, if we could all go to winecast.net everybody

00:47:55   and look at the logo in the upper left hand corner.

00:47:58   Do not make your logo look like a condom.

00:48:02   - Oh, goodness gracious. - Oh wow.

00:48:04   - That's no good. - With a tiny little sperm

00:48:06   crawling its way, bad job, Winecast.

00:48:08   - I'm also seeing a PHP warning being emitted

00:48:11   into the page on top.

00:48:12   - MySQL, GNOME Fields, expect parameter one

00:48:14   to be a resource, Boolean given.

00:48:16   (laughing)

00:48:18   - Wow.

00:48:18   - That's fine.

00:48:19   Anyway, bad logo.

00:48:21   - All right, so yeah, so it's Gary Vaynerchuk,

00:48:23   will.i.am, who I believe we had already known

00:48:26   was involved with this from a producer standpoint

00:48:29   or something, or advisor or something like that.

00:48:31   - Yeah, yeah.

00:48:32   - And like you said, Gwyneth Paltrow.

00:48:34   So I don't know, I mean, I don't think this is really meant for us, and we talked about

00:48:38   this a lot in the past, we don't need to go into this too much, but I don't think this

00:48:41   is a show meant for us.

00:48:42   I think it's, in theory, a show meant for people who are not really in the industry.

00:48:47   But I'm curious to see it.

00:48:50   I will certainly watch, just like, you know, the new Top Gear, I will watch an episode

00:48:54   or two, and, assuming, is it on Apple Music?

00:48:57   So I guess maybe I can't, because I don't subscribe to Apple Music.

00:49:00   But anyway, if I can, I will...

00:49:01   You can watch it off my Plex server.

00:49:03   Yeah, right. If I can, I will watch an episode or two. And I suspect it will not be very

00:49:09   good, but you know what, you don't know until you try. So we'll see what happens.

00:49:12   I'll watch it for Gary Vee. I like him. So I will gladly watch it for him.

00:49:16   Fair enough. All right, at this point, I think we are officially out of follow-ups. So what

00:49:20   else is awesome these days? Yeah, we need a break and a cigarette.

00:49:24   No, I don't smoke. Neither do I. None of us do. Come on.

00:49:28   I can see Jon, like, sneaking under the deck or something.

00:49:32   Smoking is not the least likely thing you'll ever see me doing.

00:49:40   If any of the three of us had a secret smoking habit, Jon would be by far the most amusing

00:49:44   one to have.

00:49:46   Oh goodness.

00:49:48   Smoking out like under the tree that drops the acorns hitting every so often.

00:49:53   I told you, the tree is defeated for the most part.

00:49:57   All the limbs are cut off.

00:49:59   Unquestionable limb when I look up might still be in line of sight, but a lot of that tree

00:50:03   is all gone now.

00:50:04   We've solved that problem.

00:50:05   That's why it's safe for my wife's new Accord.

00:50:06   Now you can get your Ferrari.

00:50:07   Not safe for a Ferrari because you don't want to leave that outside in the winter and

00:50:11   you don't want to squeeze it into my garage.

00:50:13   We are also sponsored this week by Harry's.

00:50:18   Go to harrys.com/ATP to get $5 off your first purchase.

00:50:22   You know how razor companies keep putting out new models and raising their already high

00:50:26   prices?

00:50:27   Harry's does not believe in upcharging.

00:50:30   They just made a bunch of improvements to their razors

00:50:32   and they're keeping prices exactly the same.

00:50:35   So it's still just $2 per blade cartridge

00:50:37   compared to four or more dollars you will pay

00:50:39   for the big brands at the drugstore.

00:50:41   So Harry's five blade razors now with these new improvements

00:50:43   now include, they still have the same five blades,

00:50:46   they now have a softer flex hinge

00:50:48   for a more comfortable glide,

00:50:49   they have a trimmer blade for hard to reach places,

00:50:51   they have a lubricating strip on one side,

00:50:54   and a textured handle for more control when it's wet.

00:50:56   so it had like a rubber grip on the handle now.

00:50:59   So, Harry's was founded by two friends

00:51:01   to offer people a great shave at a fair price.

00:51:04   These razors, they market them mostly towards men,

00:51:07   however, we hear from lots of women,

00:51:09   these are really unisex razors and women use them too

00:51:12   and they are great for both.

00:51:13   Now, quality's 100% guaranteed.

00:51:16   If you don't love your shave,

00:51:17   Harry's will fully refund your money.

00:51:19   And these blades are made in this incredible

00:51:22   German blade factory that Harry's bought

00:51:24   and they sell their own razors direct from this factory

00:51:28   and because of selling direct and they own the factory

00:51:30   and there's no retailers or anything else,

00:51:31   they literally charge you half the price or less

00:51:35   of what you're paying at the drugstore

00:51:36   for similar blades from big brands.

00:51:39   So I got the starter set today.

00:51:40   The Harry Starter Set is an amazing deal.

00:51:42   You get a weighted razor handle of your choice,

00:51:45   moisturizing shave cream, three precision engineered

00:51:48   five blade cartridges and a travel cover,

00:51:50   all for just 15 bucks.

00:51:52   And that's 15 bucks at the regular price,

00:51:54   But again, if you go to harrys.com/atp,

00:51:58   you will get $5 off your first purchase.

00:52:00   So that means you could get the starter kit for just 10 bucks

00:52:02   so that would be 10 bucks to cover handle, shaving cream,

00:52:06   and three blade cartridges.

00:52:07   That's incredible, that's an incredible deal.

00:52:09   So right now go to harrys.com/atp to claim that deal.

00:52:12   That's harrys.com/atp, thanks a lot.

00:52:15   - There's a rumor that Apple has pivoted their brand

00:52:23   and is going to be taking a different approach to the Apple TV, as in like the set top, the

00:52:31   television set sort of thing.

00:52:33   Again?

00:52:34   Again.

00:52:35   Recode reports that they're going to make the most baller TV guide ever.

00:52:40   That's a growth industry.

00:52:41   Yeah, right?

00:52:42   And I guess the plan is to be able to let you search and no matter where the thing you're

00:52:49   looking for may be, be it paid or not or whatever, it will figure out a way to get it to you,

00:52:56   presumably through your existing accounts at like Netflix or whatever, which I think

00:53:00   it can do already, but also perhaps by partnerships with some of the traditional TV folks, like

00:53:06   maybe ABC or CBS or something like that.

00:53:09   So I mean, this is kind of cool, and I would certainly be interested in it, but I mean,

00:53:15   Like you said, Marco, this is, I don't know, TV,

00:53:19   I don't wanna say it's not long for this world,

00:53:21   because I mean, it's been around for a long time,

00:53:22   and I don't see it going away that soon,

00:53:25   but I mean, is this the sign of major changes

00:53:29   in how we consume our television?

00:53:31   I mean, that and Netflix and Amazon

00:53:34   creating their own television shows.

00:53:38   I don't know, what do you think?

00:53:39   - This seems like kind of an extension

00:53:41   of what they're already trying to do with the Apple TV.

00:53:42   They already have Universal Search,

00:53:44   And it is, I think it is still limited to partners only,

00:53:48   on the Apple TV at least, if not on iOS.

00:53:51   So they already have like cross provider search with Siri

00:53:55   and being able to search for a TV show or movie or whatever

00:53:59   and say, all right, well it's available on Netflix,

00:54:01   on an HBO Go, and for my tune.

00:54:03   They can already list all this, that's already there.

00:54:06   So if this is just kind of an extension of that,

00:54:09   or an expansion of that, that's great, that's good.

00:54:14   I would question their ability to get these deals though

00:54:17   because it seems like we've been hearing reports

00:54:20   for what, three years now?

00:54:22   That Apple keeps trying to make a TV service,

00:54:25   like one grand new TV service to rule them all,

00:54:28   or something along those lines.

00:54:30   Honestly, there was a story I was making fun of

00:54:35   last week or whenever about Eddy Cue

00:54:38   walking into the meeting with TV executives

00:54:40   wearing shoes without socks and a Hawaiian shirt and jeans

00:54:43   whatever it was, and Apple's basically like,

00:54:47   we're Apple, screw you, method of negotiation

00:54:49   with TV companies, and who knows if that was real or not.

00:54:52   We don't really know how accurate that was.

00:54:56   - That was a story sourced from TV executives, by the way.

00:54:58   - Exactly. - Of course they're gonna say

00:55:00   the other side was unreasonable in our negotiation,

00:55:02   and they wore the wrong clothes.

00:55:03   - Exactly, so chances are, those details

00:55:07   are probably not 100% accurate,

00:55:10   but it was probably the gist of it.

00:55:12   it's very likely that that was the gist of what happened.

00:55:14   And we've heard similar attitudes,

00:55:17   we've heard of that before from both Eddy Cue and Apple.

00:55:21   So it wouldn't surprise me if the gist of this is true.

00:55:24   - I would say that the TV companies

00:55:26   are being more unreasonable in these scenarios.

00:55:29   If you are a TV company,

00:55:30   it seems like Apple's being unreasonable

00:55:31   because Apple is not budging on things they want you to do

00:55:34   that you are never gonna do.

00:55:35   But if you were to be a third party outside observer

00:55:38   and say what things are the TV people never going to do

00:55:41   what things does Apple want, you would say, "Eh, it's probably better for everyone involved,

00:55:46   or at least it's probably better for us as the consumer if TV companies who did what

00:55:50   Apple asked." But the question is, is it better for the TV companies? And maybe not, so that's

00:55:55   why we don't have a deal.

00:55:56   Well, that's the thing. I mean, we've heard for years now that Apple is, you know, they're

00:56:01   working on this new thing. And the details of what that new thing is shift slightly over

00:56:06   time, but the gist has been the same. They're working on some kind of TV plan that unifies

00:56:10   multiple TV sources, something or other, you know,

00:56:13   whether that's related to the new Apple TV or not.

00:56:15   That's been the plan for years now.

00:56:17   And it just seems like they aren't getting the deals.

00:56:20   And so maybe this approach they're taking to deal making,

00:56:24   they're walking in there as though they own the place.

00:56:28   And I think at one time, maybe they did at one time

00:56:32   have that kind of power in certain industries,

00:56:34   but I think it's pretty clear that

00:56:36   what they're doing isn't working.

00:56:37   And so when we see another report that says,

00:56:39   "Well, now the new thing is they're gonna be

00:56:42   "this version of this plan."

00:56:45   Well, show me any evidence from the past

00:56:47   that we should believe them on this.

00:56:48   Like, there's nothing.

00:56:50   - Well, that's why they're pivoting though,

00:56:51   because they've failed with the past approaches.

00:56:53   They're not sticking to the past approaches,

00:56:54   they're pivoting and trying new things.

00:56:56   But like, in some respects, I think Apple still feels

00:56:58   like time is on their side, and they may be right,

00:57:00   because the holdouts in all of this

00:57:03   are not the Netflixes of the world,

00:57:04   but it's like the ABC, NBC, Disney, you know, CBS,

00:57:08   Like, those are the difficult ones there, and all the local television, and all the

00:57:12   deals with other things.

00:57:13   So Apple's most recent pivot that we have in front of us in our houses now is like,

00:57:16   "The future of TV is apps, and we make Apple TV, and on your Apple TV you can find an app

00:57:20   for Major League Baseball, you can find an NFL app, you can find an app for ABC, CBS,

00:57:24   Netflix, Showtime, HBO.

00:57:25   Like, that's what you can find."

00:57:27   And this—I wouldn't even call this a pivot—but this latest rumor is basically that Apple's

00:57:33   new approach—sort of their version of the omnivorous box that I talked about in the

00:57:37   last show, they're like, "We will take in all content and provide you one unified interface."

00:57:41   Like the current Apple TV is like, "Here's your unified interface. It's a bunch of rounded

00:57:44   rectangles. Isn't that a great unified interface?" And you can do Siri search across them, you

00:57:49   know, like as this article says, like they already have a thing where people who make

00:57:54   the Disney app or the ESPN app or the NFL app, if you use Apple's APIs and make your

00:57:58   information searchable that when you say, "Show me whatever," that the Apple TV can

00:58:03   search across all of them. But it's not quite the same as a TV guide. Like a lot of what

00:58:06   a lot of people want to do is say, "What's on?" or "What are the series that are currently

00:58:11   running that are popular?" We can imagine a much better interface to television that

00:58:15   is independent of where the shows come from. TiVo does this, as many people have heard

00:58:19   about TiVo. TiVo has a thing where you can search for stuff and it's like, "Here's this

00:58:22   show. It's airing these episodes. These five are streaming. The season passes. Show the

00:58:25   different icons to see where you can get them from." But that's what we all want. We don't

00:58:28   want to have to switch inputs. We want one unified interface to everything the old way,

00:58:32   which Apple declined to participate in, which I still think would have been a good idea

00:58:35   because it shows so far no one has been able to do it so they could have been doing that old thing

00:58:38   anyway, is to just be a box that takes input from all the different places but nowadays TV doesn't

00:58:44   come from all these different sources from cable from these you know like it comes increasingly

00:58:49   across the internet but there's still these whole things that come across you know that all the

00:58:53   television networks as you still have the similar omnivore's box program but Apple has been able to

00:58:58   persuade the networks for the most part to put apps on their platform like we're getting close

00:59:02   guys, right? All we need to do, the last thing, is to give people like a unified

00:59:06   interface to that, to all that programming across all these apps, only

00:59:11   instead of the interface being here's a bunch of rounded rectangles, arrange them

00:59:13   how you want and pick the one you want, or talking to this terrible remote

00:59:17   control and try to find the show you want, we want to provide what people are

00:59:19   kind of used to, like a guide. Maybe it doesn't look like the big grid or

00:59:22   whatever, and we will we will be the face of television to people in the same way

00:59:25   that TiVo is the face of television to anyone who had TiVo, especially the days

00:59:28   before streaming services, but even with the streaming services it's kind of

00:59:31   worse because you've got to go find the client. But anyway, Apple wants to be the face of

00:59:34   television. The one unified interface, the one box, you never change inputs, it's always

00:59:38   on Apple TV, you can watch whatever you want. To do that, they need to make deals. The deals

00:59:43   are not happening because most of the deals are probably not in the interest of the networks.

00:59:45   And you can imagine if you're the network, like, if your only interest was, "We are the

00:59:49   network, we want to survive as a thing that can extract money," I think there's a lot

00:59:56   of problems with that long term anyway, but you would say, "Never give up the primacy

00:59:59   of, you know, never give up the interface to television to Apple. Like, don't let them

01:00:05   be the face of television. They are just decreasing your value, making you just one more source

01:00:09   of content. And then the only value you have is your ability to produce quality content

01:00:13   that people want to watch. And network television, for the most part, is terrified of that because

01:00:18   historically, they have not been really good at that. And it's only because of the legacy

01:00:22   of the fact that they have these certain frequencies in the airwaves that they're ABC, NBC, CBS,

01:00:26   you know, Fox or whatever, and although Fox arguably got its place by having quality content,

01:00:31   they're being outcompeted by HBO and Netflix and Amazon, for crying out loud in some cases,

01:00:38   in the "Hey, can we make interesting content that people want to watch?"

01:00:41   That's the only competition.

01:00:42   If it's like Apple is the interface and the programs come from these services that you

01:00:46   pay for and you pay for the service that has the shows you want, the networks are like,

01:00:51   "People are going to pay to watch NCIS colon some other word?"

01:00:54   Those people are really old and they're dying and everyone else is watching Game of Thrones

01:01:01   on streaming services.

01:01:02   And so of course they're terrified of that future.

01:01:04   But in the meantime, that basically means that Apple's strategy of like we are the unified

01:01:07   interface to all your television is a no-go because there's still enough television, particularly

01:01:12   live television, local news, and sports that is tied up behind owners and contracts for

01:01:19   companies that don't want to be basically disenfranchised.

01:01:23   So I don't know if this latest strategy of Apple is, I guess it's better than the old

01:01:27   ones and at least you can do something, but I don't know if it's ever going to work.

01:01:30   And I think the whole like, "Well, fine.

01:01:32   We don't have a deal," that Apple's trying to wait them out.

01:01:34   It's like, "Look, we don't have to do anything, network television and everything."

01:01:39   Netflix and HBO and Showtime and Amazon and AMC and all these other cable companies are

01:01:46   nibbling, coming at you from all sides.

01:01:49   They're making better content.

01:01:50   There's more of them.

01:01:51   are willing to pay for their services,

01:01:52   whereas they're only willing to pay for you

01:01:53   as part of a bundle.

01:01:54   You get to be over the air,

01:01:55   but all this legacy stuff, we'll just wait you out.

01:01:59   And so I feel like Apple is walking away from the table

01:02:00   with their flip flops and Hawaiian shirt and saying,

01:02:03   all right, well, we tried,

01:02:04   we made another run out of this time,

01:02:05   but time is on our side.

01:02:08   For every year you refuse to do a deal with us,

01:02:10   your competitors make you less and less relevant.

01:02:12   And when the generation of kids that's born today grows up,

01:02:15   they're not gonna care what the hell you are

01:02:17   and all their shows are gonna be in other networks.

01:02:19   And once that happens, we have good relationships

01:02:21   with the Netflixes of the world.

01:02:23   And they're already on our app platform,

01:02:25   and we just need to make a really good app platform.

01:02:26   They haven't quite done that yet.

01:02:28   But in the meantime, we will just

01:02:29   keep sliding that terrible remote around our rectangles

01:02:32   and being careful not to ever touch it while watching TV.

01:02:34   Don't touch it.

01:02:35   You'll mess everything up.

01:02:37   Honestly, do you think this is more about UI control

01:02:42   or more about just money?

01:02:44   If I had to take a guess, I don't

01:02:46   think the TV execs give two craps about the UI.

01:02:48   I think it all comes down to money.

01:02:50   Well, the money is the reason Apple won't do the deal because the other companies won't

01:02:55   do a deal that Apple—Apple wants a deal that's palatable to customers, and the parties

01:02:59   that have to be involved in that deal want more money.

01:03:01   Apple would have to charge too much for it.

01:03:03   That's why this never—that's basically what it comes down to.

01:03:06   So you're right, it does come down to money, but why do the networks want so much money?

01:03:10   Because it's not like the UI.

01:03:11   They just don't want to be taken out of the equation.

01:03:14   They want you to move your little rectangle to go to the ABC app.

01:03:17   They want you to know that ABC is a thing, a brand that means something, that the shows

01:03:21   are on ABC.

01:03:22   They don't want you to just say, "What's on?" and see a giant undifferentiated grid

01:03:27   with maybe an ABC logo somewhere on it and just say, "This is all television and I will

01:03:30   pick the show I want to watch," because again, that reduces them entirely to the quality

01:03:34   of the shows they produce.

01:03:35   And they don't want that because they need the other intangible BS branding stuff to

01:03:42   prop up the fact that they make worse shows than other people.

01:03:45   Like it used to be the fact that like they are channel four and they come over the airwaves

01:03:49   and they're one of the five sets of channels that come in good on your little rabbit ears

01:03:53   and therefore they have a default importance that you cannot argue with even if all their

01:03:57   shows are crap.

01:03:59   If they're just another provider of video behind a unified interface, undifferentiated,

01:04:04   made not any different than any of the other services that some of which may not even be

01:04:08   "real TV stations" like Amazon, that's not good for them.

01:04:13   So that's, you know, you're right that money, I think, is definitely a part of it and why

01:04:17   the deals don't happen.

01:04:18   But why do they want so much money?

01:04:19   What's the big deal?

01:04:21   Because if they're going to give up that role, they want to be paid for it handsomely because,

01:04:25   you know, they're CBS or whatever.

01:04:28   I feel like this is AT&T and the iPhone are singular at the time and the iPhone all over

01:04:33   again.

01:04:34   It's not exactly analogous because I think, to my recollection, AT&T are singular at the

01:04:40   time was really not doing terribly well and Verizon was just eating their lunch and I

01:04:46   think they were kind of on the ropes and knew it.

01:04:48   And to their credit, they had the wherewithal to know that they were on the ropes and make

01:04:51   this really onerous deal with Apple from their perspective.

01:04:57   But it ended up paying out for them big time.

01:05:00   And I can't help but wonder who's going to be the singular of the big American TV channels,

01:05:08   you know, the Fox, the ABC, the CBS, and NBC.

01:05:12   You know, who's gonna be the first one of them

01:05:13   to say uncle and make a deal?

01:05:16   And will that be better that way?

01:05:19   Will it be worse?

01:05:20   Will they end up becoming more popular

01:05:23   and rolling in cash?

01:05:25   Or are they just gonna be hastening their own demise?

01:05:28   - Well, and the difference here is like,

01:05:30   you only need one cell carrier.

01:05:31   - Yeah, exactly.

01:05:32   Like once you got the cell carrier,

01:05:33   now you have an iPhone as a product,

01:05:35   but until you get them all,

01:05:36   it's kind of like the music label.

01:05:37   of iTunes that the iTunes Music Store had rolled out

01:05:39   with some of the labels but not other ones is tough.

01:05:42   - Honestly, we have not seen Apple score

01:05:46   a lot of great content deals in recent years.

01:05:49   I really do question whether,

01:05:52   obviously anything we hear about these deals

01:05:55   is always rumor and speculation and everything

01:05:57   'cause they're not gonna go talk about

01:05:59   how they went or anything,

01:05:59   but it just seems like Apple's negotiating position

01:06:03   might just be wrong or possibly too arrogant

01:06:08   or asking too much or whatever the conditions are.

01:06:11   - But it depends if you think time is on their side.

01:06:13   If you think time is on their side,

01:06:14   it's like the longer we wait,

01:06:16   the next time we come to the table,

01:06:17   we will be even stronger because you will have been weakened

01:06:20   by your internet native competitors.

01:06:22   And I think that's been true.

01:06:24   Every time that Apple has gone back to the table,

01:06:25   they have been in a stronger position

01:06:27   because the networks have been in a weaker position,

01:06:28   not because of anything Apple did,

01:06:29   but because of what the competitors

01:06:31   to the networks have done.

01:06:32   So I think, I mean, you may argue if they don't make a deal,

01:06:36   then someone else will come and sweep this way.

01:06:38   But they are developing the Apple TV.

01:06:40   It is improving, unlike some other products

01:06:43   they might have where the Apple TV was in a drought

01:06:46   and now it's sort of on the track again.

01:06:49   It's just a question of whether someone else

01:06:51   is gonna get there first.

01:06:52   But nobody, like as far as I'm aware,

01:06:54   nobody has deals with all these networks.

01:06:55   Because of the iTunes thing, everyone is scared to be like,

01:06:58   as an industry, we can't all sign a deal

01:07:01   with one company because that takes away too much power. So let's all just bargain individually

01:07:05   with each things like Hulu is, I forget who's behind Hulu, is that NBC or Comcast, you know,

01:07:10   Cable Town, whatever. It's balkanized because everyone's afraid to give one technology company

01:07:15   too much power. But I don't know, I'm kind of in favor of not doing a deal that is unfavorable

01:07:24   because if they do that deal, like financially speaking, they'd have to take a loss on every

01:07:27   subscription or they would have to make it too expensive and it would be unappealing.

01:07:31   Like they have to undercut cable is what they have to do.

01:07:33   They have to be able to offer a thing that's like,

01:07:35   this is like cable, but either cheaper and way better

01:07:39   or way better and around the same price.

01:07:41   They can't say this is like cable, but 25% more,

01:07:44   but hey, there's a bunch of good features

01:07:45   'cause no one's gonna go for that.

01:07:47   - All right.

01:07:48   So Apple today has announced or let slip,

01:07:53   I'm not entirely clear what happened here, but--

01:07:55   - Announced.

01:07:56   - Oh, it was announced, okay.

01:07:57   'Cause I saw a tweet fly by of somebody taking like

01:07:59   picture of a slide at some conference or something like that, and so I wasn't sure

01:08:03   if this, like, leaked or if it was formally announced. Anyway, Apple, I guess,

01:08:07   announced that they are doing a bug bounty program, and so if you're not

01:08:11   familiar with what that is, basically that means if you find a bug in some of

01:08:17   Apple's code, and the specifics, you know, change per company, but generally

01:08:22   speaking the way it works is if you can exercise it and show Apple, rather than,

01:08:26   letting it out into the wild, if you come to Apple and say, "Hey, I found a bug. Here's

01:08:29   how you exercise it," and you do the "right thing" in quotes, then they will pay you,

01:08:36   in some cases, a tremendous amount of money for having done the right thing and brought

01:08:41   that bug to them and not just used it for nefarious purposes.

01:08:46   And what's also interesting, apparently, if you choose to donate the money that they give

01:08:51   which in some cases is up to $200,000, they will match that donation one for one.

01:08:56   So that $200,000 hypothetically becomes $400,000.

01:09:00   I think this is a great thing. The prices, the relative prices, I thought were a little bit weird.

01:09:06   Like $200,000 was for like the bootloader or something like that, or something pretty low-level,

01:09:12   which made sense. But like the secure enclave was half that or maybe even a quarter that,

01:09:18   which struck me is very weird. I would assume that the Secure Enclave, if you found a bug

01:09:21   in that, that would be worth easily as much as the highest reward, easily worth $200,000.

01:09:30   You have to price them not just how valuable it is to find the vulnerability, but how difficult,

01:09:36   which translates to how many vulnerabilities you think people will find. So it's kind of

01:09:40   depressing where you're like, what is it, sandboxing vulnerabilities is only $25,000.

01:09:44   Sandboxing vulnerabilities are serious, but I think Apple thinks there's probably a lot

01:09:47   of them and they're probably not as hard to find as boot ROM vulnerability.

01:09:51   So it's a balancing act of like how do you price these things?

01:09:54   You can't just price them on how important they are if you're kind of afraid that you

01:09:57   have like tons of sandboxing bugs because you will, you know, if it's 200K each and

01:10:02   you get 300 of them, that starts to add up.

01:10:04   Even Apple doesn't like to just give away money.

01:10:06   But the charity thing is a total Apple move to kind of like guilt you into not keeping

01:10:09   the money yourself by spending even more of their own money.

01:10:14   This has been, I think we had something way, way down the shunt so someone can find a new

01:10:17   delete later about, you know the problem with Apple is that they don't have bug bounty programs.

01:10:21   Every other company has bug bounty programs.

01:10:23   And so people find bugs in Apple stuff and it's more valuable for the people who find

01:10:27   the bugs to like sell it to jailbreak people or use it for malware than it is to go to

01:10:32   Apple because from Apple you get nothing.

01:10:33   You don't even get like a thank you.

01:10:35   Like you just throw it into a black hole and they don't fix it for a year and then you

01:10:37   fret about whether you feel — the good — the white hat people are like, "I sent you this

01:10:43   bug Apple for free.

01:10:44   It's super serious.

01:10:45   Have the suspicion that I'm not the only person in the world who knows it

01:10:48   I haven't told anybody but if I know it that probably means the bad guys know it too

01:10:52   And it's been six months since I reported it to you and I've heard nothing if I don't hear from you soon

01:10:56   I'm gonna tell the world

01:10:57   Hey guys

01:10:58   You all have a phone that's normal to this exploit and chances are good the bad guys already know about it and then Apple gets cranky

01:11:03   About that or why are you disclosing and then the people like well?

01:11:05   Why don't you fix the damn bug and people have vulnerable phones? And anyway the bug bounty program?

01:11:10   adjust the incentives to

01:11:14   To make things nicer. They have an incentive to give it to Apple. Apple has an incentive to do something about it, I suppose

01:11:19   And they you know, it's it's more likely it makes it more valuable that even the bad guys will say I found this exploit

01:11:26   How can I make the most money from it's like, you know what I can make 100k right now

01:11:31   Guaranteed if I do this and I just give it to Apple so I hope this works

01:11:35   And by the way, this was announced at the black hat conference

01:11:38   Which is this big, you know as the name implies a hacker conference or security vulnerabilities and stuff

01:11:43   As far as I'm aware, Apple has not had any formal presence or a particularly prominent

01:11:49   formal presence. Their relationship with the security community has been sort of standoffish,

01:11:55   as evidenced by not having a bug bounty program and people being cranky about sending things

01:11:59   to Apple and then not hearing anything. And this is just another—

01:12:01   JEAN-BAPTISTE. Yeah, it's about as friendly as they've been to the developer community.

01:12:04   FITZPATRICK. Well, it's a little bit worse, because when you find a security vulnerability,

01:12:08   Apple's kind of angry about it. They're not particularly grateful, and security researchers

01:12:12   done that thing where they said I sent this to you, Apple, I did responsible disclosure,

01:12:16   but you haven't done anything, so then I'm going to announce it to the world. And Apple's

01:12:18   like, don't announce it to the world, we hate you now. It's like, but I found this bug.

01:12:22   And anyway, it's a fraught relationship. But this is another tiny step along the line of

01:12:27   Tim Cook's more open Apple. The Apple sends, what was it there, their head of security

01:12:31   engineering to Black Hat to speak there, to represent Apple and to say, here we have this

01:12:37   thing that everyone else in the world has had forever. You know, they're playing catch-up,

01:12:41   But this is Apple being more open and doing more of the things that everyone has said

01:12:45   they should always do. So thumbs up.

01:12:47   Yeah, this is only good things. It is kind of embarrassing that this wasn't already in

01:12:52   place given the rest of the market. However, this is great progress and I'm glad they're

01:12:57   doing it.

01:12:58   Yep, I completely echo what you guys said. I can't believe it's taken this long, but

01:13:03   at least I got there. That's what matters.

01:13:05   One more thing on this, as someone in the chat room pointed out, as in the stories we're

01:13:08   out like, "It's Baby Steps. This is not a program where, hey, anybody who finds a bug, report it to

01:13:13   us. It's an invite-only program where you find a bug. If you are among this class of people that

01:13:19   Apple says, 'We find you are trustworthy and you have the skills, so please send us a bug.'" But

01:13:25   there's also this thing, I think this is from Gruber's site, that said, "Sources that Apple

01:13:30   mentioned, if someone outside the program discovers an exploit in one of these classes,

01:13:33   they could be added to the program." So it isn't completely closed. And I don't understand that

01:13:37   that makes sense. Look, it's closed or it's not. It's like, well, it's closed, but if

01:13:40   you find a vulnerability, we will add you to the group that's closed. So I guess if

01:13:43   you can demonstrate that you're a good guy by giving them an important bug that will

01:13:50   put you into the program, but then do you not get paid for the bug that puts you, you

01:13:53   know, apple, apple be applin', I guess. You know, they could just make it open to everybody,

01:13:59   but they don't. But it's kind of open. So anyway, maybe next year we'll be announcing

01:14:05   that black hat that the bug bounty program

01:14:07   is open to everybody, just like everyone else's

01:14:09   bug bounty program.

01:14:10   - But they would keep 30%.

01:14:12   - Yeah, right.

01:14:13   Wow.

01:14:14   - We are sponsored this week by Linode.

01:14:17   Go to linode.com/atp for a $10 credit

01:14:21   using promo code accidentalpodcast10.

01:14:24   Linode is a web host.

01:14:26   They have these Linux VPSs and I use them myself

01:14:29   and I have used them for a long time now.

01:14:32   Way before they were a sponsor.

01:14:34   I mean, they've been a sponsor for like a few months.

01:14:36   I've used them for years.

01:14:37   And I can honestly say they are my favorite web host

01:14:40   I've ever used.

01:14:41   That's why I keep using them for everything.

01:14:42   All of Overcatch is hosted there.

01:14:44   My site is hosted there.

01:14:45   You know, maybe someday ATP will be hosted there as well

01:14:48   if I ever get around to it.

01:14:48   It's a great service.

01:14:51   Plans start at just $10 a month for your own private VPS.

01:14:56   You have root access, you have full control.

01:14:59   You have so many great options.

01:15:01   They have hourly billing if you want.

01:15:04   and there's a monthly cap so you'll never pay more

01:15:06   than what the monthly rate would be.

01:15:08   They have amazing add-on services, they have backups,

01:15:11   load balancers that they call node balancers,

01:15:13   they have these long view stats,

01:15:16   and you can get a server running at Linode

01:15:17   in under a minute, it is so easy,

01:15:20   and they have the best control panel

01:15:21   I've ever used of any web host, and believe me,

01:15:23   I've used a lot of web host control panels,

01:15:25   and most of them I would not say anything nice about at all.

01:15:28   Linode is actually a pleasure to use,

01:15:30   it really is quite good.

01:15:32   You can do so many things in these things.

01:15:34   Of course you can run websites, you can run services.

01:15:36   You could also do things like run a git server.

01:15:38   If you're a developer, you wanna run

01:15:39   your own source control servers,

01:15:40   have your own control over that, you could do that too.

01:15:43   You can run VMs, you can run containers.

01:15:45   There's so much you can do at Linode.

01:15:47   It's having your own server and it's great.

01:15:48   The support is great, I've used that before.

01:15:51   I really, really enjoy Linode

01:15:52   and they're not paying me to say that.

01:15:54   They can't pay me to say that, I'm saying that anyway

01:15:55   because that's how much I enjoy it.

01:15:57   I'm so happy they're finally a sponsor.

01:15:59   They now have, and their pricing's always been good,

01:16:02   but it gets better every year or so.

01:16:05   They do hardware updates and stuff,

01:16:06   and they just make their plans better for the same prices.

01:16:09   It's so great.

01:16:10   So they now offer, for $10 a month,

01:16:13   a VPS with two gigs of RAM.

01:16:15   You can do a lot.

01:16:16   I actually have six of those doing all my feed crawling,

01:16:19   'cause for 10 bucks a month, why not?

01:16:21   Why not have six of them?

01:16:23   It's so great, so great.

01:16:25   Go to linode.com, that's L-I-N-O-D-E.com, like Linux,

01:16:28   linode.com/ATP and use promo code

01:16:32   "accidentalpodcast10" for $10 credit.

01:16:35   Thanks a lot to Linode for sponsoring our show.

01:16:37   (upbeat music)

01:16:39   - And a final note for today's episode,

01:16:43   speaking of things that have taken a long time,

01:16:46   diversity at Apple.

01:16:47   Progress, it seems, is being made.

01:16:51   I was looking at these numbers

01:16:52   when they were released a day or two ago,

01:16:55   And it didn't seem like things were too terribly rosy year

01:17:01   on year.

01:17:01   And I'm trying to find which specifically number it was,

01:17:04   but I can't at the moment.

01:17:06   But suffice to say, Apple's released their annual,

01:17:09   semi-annual diversity page and diversity numbers.

01:17:12   And one of the more impressive things

01:17:15   that I think we definitely need to applaud

01:17:16   is that they claim to have 100% pay equity across all of Apple

01:17:22   so that any job that a woman would do, that a man would do,

01:17:27   those two people should make the exact same amount of money

01:17:31   for doing that exact same job.

01:17:32   And of course, everything's open to interpretation,

01:17:35   so it's hard to say whether or not that's really real,

01:17:38   but Apple is claiming and is asserting that that's the case.

01:17:43   So that's stupendous.

01:17:47   Oh, there, here it is.

01:17:47   The number that I didn't like was data

01:17:49   from the last three years, most of the way down the page,

01:17:52   2015, 54% of Apple was white.

01:17:55   2016, 56% of Apple was white,

01:17:58   which is not getting more diverse,

01:18:01   it's getting less diverse.

01:18:03   So that's not good,

01:18:05   but I'm nitpicking perhaps on one particular data point,

01:18:10   but it's 1% less male-dominated.

01:18:14   We went from 69% to 68%, which is an improvement.

01:18:18   And one thing that they've made very clear on this site

01:18:22   is that their hiring practices are changing.

01:18:26   So it says, "Our hiring trend over the past three years,

01:18:28   we are steadily attracting more and more

01:18:30   underrepresented talent, global female new hires.

01:18:33   In 2014, it was 31%, 2016, it was 37%.

01:18:37   And U.S., what is it?

01:18:40   URM is underrepresented something or other--

01:18:44   - Underrepresented minorities.

01:18:45   - Thank you.

01:18:46   US URMs, new hires, God, URMs just sound so dismissive.

01:18:50   I don't like that at all, but anyway,

01:18:52   21% in 2014, 24%, 2015, 27% in 2016.

01:18:57   So definite improvement in new hires,

01:19:00   which should be celebrated.

01:19:02   So a little bit of good, a little bit of bad,

01:19:04   but the fact that they seem to be paying

01:19:07   this much attention to it, I think is 100% good.

01:19:10   - This is some borderline Amazon charts down here, though.

01:19:13   Like, every so-- they have three data points now.

01:19:16   So they had two, and you can make a line out of two.

01:19:19   But it's more impressive when you have three.

01:19:20   And of course, they highlight, here

01:19:21   are the lines with the slopes that are going up

01:19:23   and to the right.

01:19:24   And they are doing better with new hires

01:19:25   of underrepresented minorities.

01:19:27   They are doing better with female hires, right?

01:19:29   So they show these things, but there's

01:19:31   nothing along the y-axis.

01:19:33   There is no y-axis.

01:19:34   They just show-- they decide 21% is

01:19:36   like a centimeter from the bottom, and 27%

01:19:38   is like three times higher.

01:19:39   27 is not three times higher than 21.

01:19:41   - Oh, it's not zero based?

01:19:43   - No.

01:19:44   I mean, yeah, anyway, positive trends, progress is slow.

01:19:49   They're highlighting where they're doing the best,

01:19:51   obviously, and new hires,

01:19:52   if you're gonna do the best somewhere,

01:19:54   like that's, you know, it's, you know, forward-looking,

01:19:56   try to fix this going forward as much as possible.

01:19:59   And the other numbers, yeah, I don't know.

01:20:02   Anyway, one of the most important things

01:20:04   is that Apple has a webpage at apple.com/diversity,

01:20:08   and that they're open and transparent with these things.

01:20:10   But as always, there's a tension between, oh, good job,

01:20:15   Apple, slap on the back.

01:20:16   You really care about this.

01:20:17   Let's give you cookies for being caring and having a web page.

01:20:20   On the other hand, it's like, but on the other hand,

01:20:23   these numbers aren't awesome.

01:20:25   And so Apple's job is kind of-- it's kind of weird

01:20:27   making this web page.

01:20:28   Your job making this web page is show that Apple cares.

01:20:31   The website is there.

01:20:32   Good.

01:20:33   Show the progress Apple is making.

01:20:35   But also, be honest and upfront, as they

01:20:37   have been in the past, about where your problems are.

01:20:39   I think that's maybe where this falls down a little bit

01:20:41   because their original diversity thing was like,

01:20:43   we take a look at a diversity,

01:20:45   we are not doing a good enough job.

01:20:47   Like it was totally unflinching saying like,

01:20:49   oh, there's good and bad.

01:20:50   Now the first run of this, I forget what year it was,

01:20:52   was like, we are not happy with this.

01:20:54   We are not doing a good job.

01:20:56   We are not meeting our own standards for how this should be.

01:20:59   And the same pages, this page this year is more about like,

01:21:02   hey, we're doing well and everything.

01:21:04   I'm sure internally they still have their eyes on the prize

01:21:06   and like, all right, there's still progress to be made,

01:21:08   But there is a danger of falling into the trap

01:21:12   where every time they've come out with these numbers

01:21:14   that we just parrot back the cherry stats

01:21:18   that are getting better and don't realize

01:21:19   that the overall picture is still pretty grim.

01:21:22   So I don't know, I don't wanna slam them

01:21:24   for not making progress faster because again,

01:21:27   it takes a long time to turn a ship this big.

01:21:30   It's not like you're gonna fire all your employees

01:21:32   and start over again from scratch.

01:21:34   And new hiring is the place where you can fix things

01:21:36   and they're doing better.

01:21:37   But on the other hand, there's a long road ahead.

01:21:40   So I hope this page is still here 15 years from now.

01:21:42   And I hope if you were to do the 15 year graph

01:21:44   with an actual labeled Y axis,

01:21:46   that it would still show equally encouraging,

01:21:49   a zero based Y axis,

01:21:50   it would show equally encouraging trends.

01:21:53   - Yeah, it's tough because I want to celebrate this so much,

01:21:58   but there's a lot of room for improvement.

01:22:02   I mean, 2016 global gender in what they describe as tech

01:22:06   is 77%, and 56% of it is white.

01:22:09   Like, that's a lot of room for improvement there.

01:22:12   But I mean, if you look at-- to be fair,

01:22:15   if you look at my company, there are four Android developers,

01:22:18   four iOS developers, and every single one of us

01:22:21   is a white male.

01:22:22   So I mean, I shouldn't really be throwing stones on this myself.

01:22:25   But I hope we get better.

01:22:29   I would love to see us get better.

01:22:30   And hopefully, as we open other offices, we will get better.

01:22:34   But it's, I don't know, it's hard and it shouldn't be, but it is.

01:22:40   Thanks so much for our three sponsors this week, Hover, Harry's, and Linode, and we

01:22:45   will see you next week.

01:22:46   Now the show is over, they didn't even mean to begin, 'cause it was accidental, oh it

01:22:57   was accidental.

01:22:58   John didn't do any research, Marco and Casey wouldn't let him.

01:23:04   'Cause it was accidental (it was accidental)

01:23:07   It was accidental (accidental)

01:23:10   And you can find the show notes at ATP.fm

01:23:15   And if you're into Twitter, you can follow them

01:23:20   @C-A-S-E-Y-L-I-S-S

01:23:24   So that's Casey Liss M-A-R-C-O-A-R-M

01:23:28   ♪ M-A-N-T-M-A-R-C-O-R-M-N-S-I-R-A-C ♪

01:23:33   ♪ U-S-A-C-R-A-C-U-S-A ♪

01:23:36   ♪ It's accidental ♪

01:23:38   ♪ It's accidental ♪

01:23:39   ♪ They didn't mean to ♪

01:23:42   ♪ Accidental ♪

01:23:43   ♪ Accidental ♪

01:23:44   ♪ Tech podcast ♪

01:23:46   ♪ So long ♪

01:23:48   - All right, so Marco, you and I went through

01:23:51   a deeply painful episode last week

01:23:55   when we had to hear about TiVo for entirely too long.

01:23:58   - And remind me again why that was so painful?

01:24:00   - Because I just don't care.

01:24:01   I just really don't care.

01:24:02   - It's painful to hear about things you don't care about?

01:24:05   It's painful?

01:24:06   - No, I'm just being silly.

01:24:07   It's not painful.

01:24:09   - I actually, I thought it was gonna be worse than it was.

01:24:11   I just zoned out and I just pretended

01:24:14   like I was listening to Hypercritical.

01:24:16   And I just forgot that I could talk for most of the time

01:24:19   and just pretend like I was listening to Hypercritical.

01:24:21   And for that it was great,

01:24:22   because it was basically a brief interlude of Hypercritical.

01:24:26   - Yeah, that's fair.

01:24:27   - No, I'm just giving you a hard time, Jon.

01:24:30   - Well, this is not gonna be a brief interlude

01:24:31   of build and analyze, because I don't think

01:24:33   you would have ever talked about the MP3 file format

01:24:35   on build and analyze.

01:24:36   - No, I didn't know that much about it back then.

01:24:38   - But even if you did, this is very nitty gritty,

01:24:40   I feel like.

01:24:41   - Yeah, so tell us what it is that we need to know

01:24:43   and don't know about the MP3 file format,

01:24:45   because I genuinely am very interested.

01:24:47   - So this is basically kind of like this developer rat hole

01:24:50   I fell into last weekend, where I had some work time,

01:24:53   and rather than do what I'm supposed to be doing this summer,

01:24:56   just updating Overcast to iOS 10

01:24:58   and making a whole new watch app

01:24:59   and making a today widget and all this other garbage.

01:25:01   And today I was procrastinating by working on forecast,

01:25:04   my MP3 encoder.

01:25:05   And I decided, you know, let me just do

01:25:08   whatever work is required to make it work with VBR output.

01:25:12   So to back up a little bit,

01:25:14   just a high level version of the MP3 file format.

01:25:19   So the way MP3s work at a very high level,

01:25:22   and please, if you're a nerd about this stuff,

01:25:24   please forgive me about these details

01:25:25   if I'm getting any of them wrong.

01:25:26   I'm trying to give a very high level overview here.

01:25:28   The way lossy compression works is basically

01:25:31   try to not store things that you probably won't notice

01:25:36   if they're not stored.

01:25:37   And then for the things you do notice,

01:25:40   try to store them with less precision.

01:25:42   In an MP3, one of the famous ways it does this

01:25:46   is by omitting sounds that you probably won't hear.

01:25:49   And so obviously things that are outside

01:25:51   the range of human hearing, that's the easy one.

01:25:53   They also do things like, there's a principle

01:25:55   called masking, where if there's a very quiet sound

01:25:59   and a very loud sound at the same time,

01:26:01   the very loud sound is going to be so overpowering,

01:26:04   the very quiet sound is just gonna be drowned out.

01:26:06   So there's no reason to store the information

01:26:09   about the very quiet sound,

01:26:10   because the very loud sound, that's all you'll hear.

01:26:12   One way they achieve the size savings

01:26:14   is just by omitting things that are just kinda drowned out

01:26:17   or that you won't hear.

01:26:19   Another way they do it is by reducing the precision

01:26:21   of the things you do hear.

01:26:22   And so there are a few different tricks they can do

01:26:25   for this 'cause basically the precision at which

01:26:28   we perceive what we hear is not constant

01:26:31   throughout the frequency range.

01:26:32   Very low frequencies, very high frequencies,

01:26:35   we tend not to have as much precision

01:26:38   about perceiving those things.

01:26:39   And so they can store those less precisely

01:26:42   and therefore using less data.

01:26:44   They can also do things like in a stereo recording

01:26:47   where you might have very, very similar sound

01:26:50   coming out of left and right channels

01:26:51   with just slight differences.

01:26:53   So there's a method called joint stereo

01:26:55   where basically this is the left channel

01:26:57   and the right channel differs by this much

01:26:59   and just sort of the difference for the right channel.

01:27:01   We are also very bad at perceiving not only the details

01:27:05   about very high-pitched and very low-pitched sounds,

01:27:07   but also where they're coming from.

01:27:09   And you might realize this,

01:27:10   like if you can hear a very high-pitched sound

01:27:12   like in your house, like back in the old days,

01:27:14   like you'd hear the very high-pitched whine of a CRT TV

01:27:17   and you could just walk into a house,

01:27:19   you can hear, like I can hear there's a TV on

01:27:21   somewhere in the house,

01:27:23   but you might not be able to pick out where exactly,

01:27:25   like what direction exactly it was coming from.

01:27:27   Also, similar reason why subwoofers in home theater systems

01:27:31   tend to just be one subwoofer that you just put somewhere,

01:27:34   and it kinda doesn't matter,

01:27:36   because the very low frequencies, again,

01:27:37   you're not nearly as good at perceiving

01:27:39   where they're coming from.

01:27:41   So they can do things there too,

01:27:42   they can save space there too, with things like,

01:27:44   all right, well, you know,

01:27:45   if we have to store this separated stereo image here,

01:27:49   maybe we can just store the average of the very high

01:27:51   and very low stuff in the middle

01:27:53   and not have to worry about the sides,

01:27:54   you know, not have the separation.

01:27:56   So the whole principle of the MP3 file format

01:27:59   and all lossy audio formats is based on this idea

01:28:03   of like figure out what we can either omit entirely

01:28:07   from storing and then figure out tricks we can use

01:28:10   to store it less precisely.

01:28:12   Obviously though, as you lower the amount of space

01:28:15   you're willing to spend on it, as you lower the bitrate,

01:28:17   how many bits per second you're willing to devote

01:28:19   to storing this, you start hearing artifacts,

01:28:23   you start hearing the quality loss,

01:28:24   you start hearing, oh, now this is sending muffled

01:28:27   or that's sending distorted or that's sounding weird

01:28:29   or that symbol hit kinda sounded weirdly telephonic.

01:28:33   You start hearing flaws.

01:28:36   I don't wanna get into too many of the details

01:28:38   of the argument over whether you can hear

01:28:39   the difference or not.

01:28:41   Generally, most tests show that about 192 kilobits

01:28:45   per second, you don't really hear the difference

01:28:47   in most things for most people.

01:28:49   That's beside the point though.

01:28:50   So when you're encoding a podcast,

01:28:52   there's a few different ways you can go about

01:28:55   managing the bitrate, how many bits per second

01:28:58   you are willing to spend on the audio.

01:29:00   The most direct kind is constant bitrate or CBR,

01:29:05   which basically, so MP3 files are divided into frames,

01:29:08   and it's just a time slice, every frame is 1152 samples.

01:29:13   whatever your sample rate is, like at 44K,

01:29:16   that's like 26 milliseconds,

01:29:18   every frame you have a bit rate and you say,

01:29:20   all right, well in constant bit rate mode,

01:29:22   every frame will get 96 kilobits or 64 kilobits

01:29:26   or 128 kilobits or whatever.

01:29:28   And that's a very simple way to do things

01:29:30   and that mostly works and podcasts

01:29:32   are almost always encoded that way.

01:29:36   I started to wonder why exactly,

01:29:38   because we also have these other methods

01:29:41   that are based on variable bit rates, or VBR.

01:29:45   The encoder has some idea about the complexity

01:29:48   of each frame, each one of those little 26 millisecond,

01:29:51   each one of those little time slices.

01:29:52   The encoder can decide, this part I'm encoding right now,

01:29:56   this little time slice, is a pretty complex,

01:29:58   there's a lot going on here, so to encode this

01:30:01   with a certain degree of perceived quality,

01:30:03   I need more bits.

01:30:05   And then maybe like three seconds later,

01:30:07   there's a quieter passage, or a simpler passage,

01:30:11   and you can say, you know, this part,

01:30:13   I don't need this many bits,

01:30:14   I can encode this at a lower bit rate.

01:30:15   You know, you can have the encoder kind of decide

01:30:18   on a target perceived quality level

01:30:20   and just use as many bits as you need

01:30:22   to achieve that quality level

01:30:23   and have it just vary constantly throughout the file.

01:30:25   For podcasts, this is an obvious choice, right?

01:30:29   It makes sense that for podcasts,

01:30:31   that should work really well

01:30:32   because podcasts have a lot of silence.

01:30:35   This is, I've kind of built my current living on this.

01:30:37   They, podcasts have a lot of silence

01:30:40   and voice is pretty easy,

01:30:42   and then occasionally you throw in like a music clip

01:30:44   or a music bed running under things,

01:30:46   or a theme song, or a clip from TV or something like,

01:30:50   so you occasionally have more complex stuff

01:30:53   that could use more complexity.

01:30:54   And like in our show, we started out being

01:30:57   a 64 kilobit mono show for like the first year or two.

01:31:02   And it sounded okay, it didn't sound great,

01:31:05   it sounded okay.

01:31:05   One of the things that sounded the worst

01:31:08   was our theme song because 64K mono is kind of terrible for music. A while back now, maybe

01:31:14   a year ago or something like that, I switched to 96K stereo and it made the theme song sound

01:31:21   way better and anytime we'd insert a clip like from a Steve Jobs keynote or anything

01:31:25   it made all those sound way better, any kind of musical clip or any kind of insert sounded

01:31:28   way better and our voices sounded better too. And even though the entire rest of the podcast

01:31:35   is us talking and I don't do any kind of stereo separation,

01:31:38   so it's just a mono thing.

01:31:40   Because of the way the joint stereo encoding works,

01:31:42   we just get all 96 kilobits to our mono channel

01:31:46   for the entire rest of the file.

01:31:48   Because it can say, all right, well,

01:31:50   the channels are encoded as like, you know,

01:31:52   the main channel equals this,

01:31:53   and the difference in the other channel is zero, basically.

01:31:56   So you get all the bits to yourself.

01:31:58   And then only when you have a stereo thing inserted

01:32:01   do you then split up the bits

01:32:03   as necessary between the channels.

01:32:04   So it is really a great way to do it,

01:32:06   but what would be even better would be a VBR encoding.

01:32:10   All these silences between all the words I'm speaking,

01:32:13   those would get like the minimum frame size,

01:32:15   which I think is 32 kilobits per second.

01:32:17   So like the minimum frame size for all those silences,

01:32:20   'cause it doesn't really matter,

01:32:20   you won't hear the difference,

01:32:22   and then when we throw in a music clip or something,

01:32:23   that could go all the way up to like 192

01:32:25   to really get the music to be perfect quality.

01:32:28   If you did it that way,

01:32:29   it wouldn't take up very much more space.

01:32:31   In fact, it probably takes up less space,

01:32:33   and in my tests, it actually would take up about

01:32:35   maybe 25% less space than my constant 96K

01:32:40   to have similar voice quality as we have now,

01:32:43   but then have the ability to put in our theme song

01:32:46   at effectively perfect quality.

01:32:48   So why don't we do this?

01:32:50   So I spent the weekend adding the capability to forecast

01:32:53   to say, you know what, let me just give it the ability

01:32:55   to output VBR files, 'cause it wasn't that much more work,

01:32:57   and I got to dive into the format

01:32:59   and learn a bit more about it.

01:33:00   There's also, for completeness,

01:33:02   there's something called AVR, which is average bit rate.

01:33:05   And the idea here is, it is VBR,

01:33:09   but instead of targeting a certain quality level,

01:33:12   it just says try to keep the bit rate

01:33:13   at exactly this average over time.

01:33:17   So basically if you have like a couple of very brief frames

01:33:22   where you can like, you know, for this second of audio,

01:33:26   you need more quality,

01:33:27   but for the other 30 seconds around it, you don't.

01:33:29   You know, you can have little temporary jumps there,

01:33:31   But ABR would not work for the case I'm talking about,

01:33:35   which is if I say average bit rate of our whole file

01:33:39   it needs to be this,

01:33:40   well the theme song is gonna just blow that

01:33:42   because the theme song needs like two minutes

01:33:44   of really high quality,

01:33:46   so the average during that time is gonna be way higher.

01:33:48   And so it basically doesn't work right.

01:33:51   Like you could have a few seconds of higher quality

01:33:53   but not minutes of higher quality.

01:33:55   So that wouldn't work for our needs.

01:33:58   So I really started trying to figure out

01:34:00   Like, how can I get a true VBR encoding

01:34:04   in the world of podcasts?

01:34:05   'Cause again, VBR has been around for almost 20 years.

01:34:08   This is not a new thing,

01:34:10   and yet almost no podcasts or VBR.

01:34:12   Why?

01:34:13   Off the top of your head,

01:34:14   can you guys think of why this might be?

01:34:16   - Because podcast, podcatchers don't support it

01:34:20   for some reason or another,

01:34:21   or they didn't at some point.

01:34:23   Or what about like the hardware, actually?

01:34:25   The old hardware, the old iPod hardware?

01:34:28   - Good question.

01:34:28   Honestly, I think by the time iPods came out in 2001,

01:34:31   I think all the hardware supported it.

01:34:33   Like in the very early days,

01:34:35   some hardware would have problems with it,

01:34:36   and maybe if you had one of the first MP3 CD players

01:34:40   or MP3 Flash players, like the Diamond Real,

01:34:42   if you had some of the very first MP3 players

01:34:45   or software or car stereos that played MP3s,

01:34:48   maybe there'd be a problem there.

01:34:50   But VBR compatibility has been solved so long ago

01:34:54   in all this stuff,

01:34:56   because it's literally almost 20 years old.

01:34:58   I was like, "My music has always been VBR.

01:35:00   I've never done CBR.

01:35:01   From the second I ever made an MP3, I had the choice, VBR or Constant Bitrate, and it

01:35:06   was like, why would I choose Constant VBR always?"

01:35:09   And it's always played, and obviously I started listening to it on—actually, I did.

01:35:12   We had a Yamaha MP3 player for, like, running or whatever, like the size of a shuffle, and

01:35:18   that played VBR, so I don't think it's the hardware either.

01:35:21   - Yeah, so what I found out, the main problem with VBR

01:35:26   is streaming.

01:35:28   When you're streaming, when a player plays back

01:35:31   a stream file, if you need to jump ahead to a timestamp

01:35:36   and you haven't downloaded that part of the file yet,

01:35:38   like you don't have that far, the way this is usually done

01:35:41   is the player will download the first few,

01:35:46   100 kilobytes maybe, basically they'll download

01:35:48   the first part of the file to get all the header

01:35:49   information and everything, all the metadata,

01:35:52   and then they will terminate that connection

01:35:54   and make a new connection that jumps ahead

01:35:56   using a range request to begin playback

01:36:00   like 50 megabytes into the file.

01:36:03   So it doesn't have to download everything

01:36:06   in the middle there to get there.

01:36:08   So the problem is it needs to be able to predict

01:36:10   at what byte offset in the file maps

01:36:13   to the timestamp that it's going to.

01:36:16   It also needs to know how long the file is.

01:36:17   duration is another challenge here.

01:36:19   And with a constant bit rate or AVR,

01:36:24   average bit rate kind of scheme,

01:36:26   you can do that pretty effectively.

01:36:28   MP3, once you have the byte stream,

01:36:32   I mentioned this in the past show,

01:36:33   once you have the byte stream,

01:36:34   you can jump ahead to a certain byte point,

01:36:36   and then every MP3 frame,

01:36:38   every one of those 26 millisecond time slices,

01:36:40   begins with a certain byte pattern

01:36:42   that's easy to seek to and locate.

01:36:44   So you can jump into an MP3 byte stream

01:36:46   any point, any byte, and you can just scan forward

01:36:49   until you see 11 ones, basically,

01:36:51   an FFE or whatever it is.

01:36:53   You can scan forward until you see that,

01:36:55   and then that's your frame header,

01:36:56   and you can start playing from there.

01:36:57   But you still have to know where you are.

01:37:00   So if you jump ahead to byte position,

01:37:02   you know, 50 megabytes, expecting that to be time stamp

01:37:06   one hour and 20 minutes, there's nothing in the file,

01:37:09   in the byte stream, that says,

01:37:11   I am time stamp one hour and 20 minutes at that point.

01:37:15   So you have to already know the timestamp that you are at.

01:37:18   You have to keep track yourself as the decoder,

01:37:20   as the player.

01:37:21   And so in a file where you know the constant bit rate,

01:37:24   where it's kept the same, you can just do the math.

01:37:27   You can say, all right, well, I know the music data began

01:37:29   at byte zero, and you know the file is 100 megabytes,

01:37:33   and you know the duration from the header

01:37:34   says it's an hour long.

01:37:36   So if you jump to 0.50 megabytes,

01:37:37   that's right in the middle, so that should be 30 minutes.

01:37:39   Done.

01:37:40   In a constant bit rate file, that's true.

01:37:42   you know, probably solve this problem.

01:37:45   QuickTime.

01:37:46   Mm.

01:37:47   I know you hate those container formats and the MPEG-4

01:37:51   for container formats that spawn from it

01:37:53   because they're all complicated and you have multiple atoms

01:37:55   and streams or whatever.

01:37:56   But I'm pretty sure they solve this one.

01:37:58   But anyway.

01:37:59   Yeah, QuickTime has a number of other problems, actually.

01:38:01   But anyway, so--

01:38:02   The fact that Apple is not interested in it

01:38:04   anymore being the most primary one.

01:38:06   But anyway, these exact problems,

01:38:08   like to be able to pick different codecs,

01:38:09   have them be very low bit rate to be able to have time codes

01:38:12   and other, you know, multiple streams to tell you how,

01:38:16   I've come to a point in the file,

01:38:18   how far am I in the file and what is the subtitle

01:38:20   I should be showing and what is the picked image

01:38:23   that I should be displaying on top of the, anyway.

01:38:25   - Yeah, one of the fun challenges

01:38:27   about the wonderful QuickTime file format

01:38:29   and its undocumented chapter spec

01:38:31   is that in MP3 chapters,

01:38:34   all the chapter info is right up front in the file.

01:38:36   So you can read like the first couple hundred kilobytes

01:38:38   and have all the information you need

01:38:39   to show the entire table of contents,

01:38:41   and then you can jump to the point you need.

01:38:42   - QuickTime predates internet streaming,

01:38:44   QuickTime predates internet.

01:38:46   - Yeah, but QuickTime chapters doesn't,

01:38:47   and they still did it this way.

01:38:48   So in the QuickTime format, important information

01:38:52   like the chapter titles are spread

01:38:55   throughout the entire file.

01:38:57   Like the title occurs in the file when the music does.

01:39:00   Like in the audio at that point,

01:39:02   the title's interleaved there.

01:39:04   So in order to display the table of contents,

01:39:07   you have to have the entire file, basically.

01:39:10   So that's a bad design for this kind of use.

01:39:12   Anyway, the main problem with the Ember 3 format

01:39:14   is with seeking and duration estimates

01:39:18   and streaming of EBR files

01:39:22   is that you need to know what byte position in the file

01:39:27   maps to what timestamp.

01:39:29   And they figured this out early on.

01:39:31   This was a problem right from the start.

01:39:33   People in the chat are saying that they used to have

01:39:35   maybe old software or old hardware

01:39:37   that would display the wrong duration on VBR files.

01:39:40   One of the ways decoders would do this

01:39:42   would be to just read the first couple MP3 frames

01:39:46   and figure out the average bit rate of those frames

01:39:49   or even just read the very first one

01:39:51   and then just look at the file size and say,

01:39:53   "All right, well, we're gonna assume

01:39:55   "this represents the average bit rate of the file

01:39:57   "and extrapolate from the file size how long this file is."

01:40:00   And that's dumb and doesn't work.

01:40:01   So that's why those programs would often display

01:40:03   the wrong durations.

01:40:04   There is also an ID3 tag value of the duration of the file,

01:40:09   but not everything supports that,

01:40:11   not every encoder embeds that,

01:40:12   and somebody might have edited that, so it might be wrong.

01:40:15   Early on, they figured out a little solution to this problem

01:40:19   and it's, do you guys remember back when people would argue

01:40:22   about MP3 encoders, do you remember the encoder

01:40:24   that was Xing or X-ing, it's X-I-N-G?

01:40:27   - No.

01:40:28   - Anyway, they figured this out early on.

01:40:30   The current hack to do this is in these MP3 frames,

01:40:34   which are like 300 bytes long for this kind of file.

01:40:38   In these frames, the very, very first audio frame

01:40:43   in the file is called an info frame for VBR files.

01:40:47   And they basically write all zeros of the audio data

01:40:50   and then they have a bunch of free space in the frame

01:40:52   'cause they didn't use it all.

01:40:54   So they have this special format where they embed

01:40:58   really, really tiny metadata and one of the things

01:41:01   they embed is a 100 byte seek table

01:41:05   that literally just maps percentage points

01:41:08   to the unsigned character value,

01:41:11   so you have 255 values there,

01:41:13   of like, it maps the duration percentage

01:41:17   to the byte percentage of the file.

01:41:19   - That's fascinating.

01:41:21   - Yeah. (laughs)

01:41:23   - I'm sure there'll be no rounding errors

01:41:25   with that long files, but that's--

01:41:26   (laughing)

01:41:27   - Yeah, also incredibly imprecise, right?

01:41:29   Like for a song, if the song is more than a minute

01:41:32   and a half long, you already don't have

01:41:35   second level precision.

01:41:36   You already are like less than one second precision.

01:41:39   For a podcast, that's like less than one minute precision.

01:41:42   That's even worse.

01:41:44   So that is terrible, right?

01:41:47   And that, it turns out, that is for the most part

01:41:51   what most Apple playback interfaces,

01:41:55   you know, most of the APIs, the AD player and everything,

01:41:57   That is what most of these things will use,

01:41:59   and this, again, this info frame with this jump information

01:42:02   and it has been around for a very long time.

01:42:04   So that's what most hardware will use with VBR files

01:42:08   to just be able to tell, like, all right, well,

01:42:09   this VBR song, if you seek ahead to point X

01:42:12   and we don't have the whole file,

01:42:14   we know we can jump to about this byte position

01:42:16   and be approximately correct within a couple of seconds

01:42:19   for a three-minute song.

01:42:20   Doesn't really work for podcasts, right?

01:42:23   So the idea I had, I'm making the encoder,

01:42:26   I make the player, what if I just make,

01:42:30   I define a new ID3 tag that gives way more,

01:42:34   I could do second level precision

01:42:36   and just have it be as long as it needs to be,

01:42:38   or something like that.

01:42:39   Whatever it is, I come up with a scheme

01:42:41   that is the size wouldn't matter

01:42:43   for 100 megabyte podcast file or whatever,

01:42:46   'cause it could be like 15K

01:42:47   and have all the information I would need.

01:42:49   So I figured, I was drafting this plan in my head

01:42:52   of what if I just do this?

01:42:55   And the main problem with this is,

01:42:58   even if Overcast supports it,

01:43:01   nobody else would support it,

01:43:03   because how many people do you think are working on

01:43:05   the low-level MP3 decoding libraries at Apple or Google?

01:43:09   This is ancient stuff now.

01:43:11   It's like, so many people have tried to modify

01:43:14   and advance the JPEG format,

01:43:16   and none of them ever take off,

01:43:18   because nobody is still working on their JPEG decoders.

01:43:22   Like, there is no new version of JPEG

01:43:24   that's going to ever matter because we have JPEG already

01:43:26   and that's everywhere and nobody wants to touch it

01:43:28   'cause they consider it a solved problem.

01:43:29   MP3's the same way.

01:43:31   There are other audio formats and advancements

01:43:33   and everything and most of them have really gone

01:43:35   effectively nowhere with the exception of AAC

01:43:38   'cause Apple uses it everywhere, but for the most part,

01:43:41   most improvements have gone very, very few places

01:43:44   because basically nothing implements them

01:43:47   and no one cares, right?

01:43:49   - So if you fed a VBR file to one of these

01:43:51   non-overcast players and they ignore your ID3 tag because they have no idea what it

01:43:55   means, what would they do for duration and skipping around? Like they would just read

01:44:02   that little zing thing if it was present and that's it? Like I don't understand how they

01:44:06   could even use that zing thing but it's like again I jump to this offset, does it just

01:44:09   display the exact number of seconds that it should be according to its math with some

01:44:13   rounding and then just be like that's not the real offset but oh well that's the best

01:44:17   we can tell you if you're streaming.

01:44:20   - That literally is what happens.

01:44:21   If you have the whole file,

01:44:23   the file you can just scan forward

01:44:26   and scanning forward is incredibly fast

01:44:29   because you're dealing with very small data ranges here

01:44:33   and in order to find and read an MP3 header

01:44:38   is incredibly simple bit shifting.

01:44:40   It's very, very simple stuff

01:44:42   because this is an old format

01:44:43   designed for really slow computers.

01:44:45   So if you have the whole file,

01:44:47   you can seek back and forth just by reading all the frames

01:44:49   and keeping track yourself.

01:44:51   And you can have perfect accuracy there.

01:44:53   It's only an issue with streaming.

01:44:54   And only an issue if you are,

01:44:56   and streaming, if you're playing from the beginning,

01:44:58   it isn't a problem.

01:44:59   But it is a problem if you're trying to jump ahead

01:45:01   to a timestamp where you have not downloaded

01:45:03   the intermediate part of the file

01:45:05   between the beginning and that timestamp.

01:45:07   That is the only place this is a problem.

01:45:09   - The obvious terrible solution that springs to mind for me

01:45:12   is all right, fine, then Overcast just makes

01:45:15   a different request to the server and if it has a VBR version with a special thing it

01:45:20   serves it and if it doesn't, you know what I mean?

01:45:21   Like put all the smarts in the – it's terrible, I know, but it would totally work.

01:45:26   Because you're like everyone else would get the normal constant bitrate one and then now

01:45:29   you'd have to make two versions of ADP, one special one for – one overcast savvy version

01:45:33   that would be VBR and better quality and blah, blah, blah.

01:45:37   But everyone else would get the other version and that's a terrible solution because you

01:45:40   haven't changed what anyone else does but you have made your player slightly better

01:45:42   but now you have to encode everything twice.

01:45:44   And if anybody else wanted to do that,

01:45:46   which I assume you'd want other people to do it too,

01:45:48   they'd be pissed at you because now you have to

01:45:49   encode everything twice and it's dumb.

01:45:50   - Yeah, and there's all sorts of other problems with that.

01:45:52   For example, we'd have to leave Squarespace.

01:45:55   Because here's the thing,

01:45:56   podcasts have been around for so long

01:45:58   that there's all these WordPress plugins

01:46:00   and CMSs like Squarespace that have podcast support.

01:46:04   But I don't think any of them have the ability

01:46:07   to have like, oh, you know what, in my feed,

01:46:10   actually every entry's now gonna have two enclosure tags.

01:46:12   You just do a stupid can you do a convention over configuration dot mp3 dot Marco's weird version? Oh god

01:46:19   That's even worse

01:46:20   Like the RSS feed would just say that mp3 but overcast would know actually make a request for that

01:46:25   Mb3 dot Marco's weird version first if you get a 404 then make a request for the yes

01:46:29   I'm telling you this is a terrible like this is the the obvious terrible solution that comes to mind immediately

01:46:34   That you should probably ever do but like people have done worse things like the other one is the whole embrace and accent thing

01:46:40   "Oh, get it added to the ID3 spec, socialize it,"

01:46:43   something like that, that's how everything happens.

01:46:44   And you can make it de facto standard,

01:46:46   I'm just not sure you have the market share

01:46:47   to pull that off at this point.

01:46:48   - Yeah, and the other problem is like,

01:46:50   you know, if I actually made just the regular file VBR,

01:46:55   one of the biggest problems here is my timestamp share links

01:46:59   because when you open up a timestamp share link

01:47:03   that Overcast generates for, you know,

01:47:04   share this podcast at this point in time,

01:47:07   it's using the HTML5 audio tag.

01:47:09   And that's just Apple's decoder.

01:47:11   It's gonna load up, it's gonna use Apple's decoder,

01:47:13   and I've tested this with VBR files,

01:47:15   and it's just off, it doesn't work correctly,

01:47:19   it does not seek correctly.

01:47:20   - Yeah, the share link would have to use the .mp3

01:47:24   instead of the .mp3 .MarkosWeirdVersion.

01:47:26   - Yeah, I mean, basically the only way this works

01:47:29   is the MarkosWeirdVersion, but all of it,

01:47:32   the truly sad part about all this is

01:47:35   after doing all this research,

01:47:36   after figuring out this crazy info frame

01:47:38   the rest of the VBR file format.

01:47:40   Now I have this awesome parallel VBR encoder

01:47:42   that I basically can't use.

01:47:44   'Cause even if I did the craziness required

01:47:48   to make this work with Overcast,

01:47:50   ATP would probably be the only podcast that ever did it.

01:47:53   Because most podcast producers simply don't care

01:47:57   about audio quality very much.

01:47:58   To them, they encode it at like 64K and that's good enough.

01:48:01   And maybe they're paying per gigabyte

01:48:03   and so maybe they can't afford larger files.

01:48:05   Funny thing there though is like,

01:48:06   Even if you were doing 64K mono, I've done tests on that too.

01:48:09   VBR would save them like 25-30% for most shows, but most people are not interested in causing

01:48:15   a possible headache with certain players in exchange for a 30% file savings.

01:48:20   It would be like handcrafted artisanal podcasts where like, "Yeah, only you would do it,"

01:48:26   and five other people in Brooklyn would do it, but it's like, "Oh, well, everyone

01:48:29   knows you have to encode it twice."

01:48:30   One .mb3 for the peons, and one .mb3 .marco's weird version, which by the way is bad branding.

01:48:35   If you come up with a clever name for it instead of .mp3, it would be like .mpz, which is probably

01:48:40   already taken, or some other .mp3, .something else.

01:48:44   You could brand this in a way that it's like, yeah, nobody does this, but the people who

01:48:47   really care about it, people who really care about locally sourced, handmade, fair trade

01:48:53   podcasts, they encode everything twice.

01:48:58   And the one good player that cares about it always makes a request for the .mpz file first.

01:49:03   and if it 404s, it requests the .mp3,

01:49:06   but if it doesn't, it plays the .mpz and it's better.

01:49:09   - Yeah, so basically I went on this giant expedition.

01:49:13   I achieved a lot, I made forecast a lot better

01:49:15   by making it write that crazy info frame

01:49:18   and understand the format a lot better

01:49:20   and be able to do VBR if it ever needs to.

01:49:22   But the moral of the story is

01:49:24   I ran into a whole bunch of barriers

01:49:26   that basically nobody will ever care as much as I do to fix

01:49:30   and that make it pretty much impossible

01:49:31   to really use for podcasts in a responsible way.

01:49:34   - Here's the other angle on it.

01:49:37   Remember when Microsoft had secret APIs

01:49:39   that only they could use to make their apps faster

01:49:41   and everything and people were all angry about it?

01:49:43   So you could do this, frame this as like,

01:49:45   this is a secret overcast API that only ATP,

01:49:48   'cause everyone is always obsessed with the idea

01:49:49   that ATP is like a preferential treatment

01:49:51   in your podcast app or whatever.

01:49:53   But like, that only ATP has access to it and you do it,

01:49:57   and then when someone comes to you and say,

01:49:58   well actually, it's not a secret API,

01:49:59   here's a webpage that's been up for a year

01:50:01   telling you if you want to do this, make a .mpz file and you can do it.

01:50:04   And then suddenly it's on them to be like, if they come back and say, "Oh, well that's

01:50:08   annoying."

01:50:09   It's like, "Well, you wanted the secret API."

01:50:11   You're wondering, "Why does overcast, why does ATP sound so good?"

01:50:14   And the file size is so small and the other podcasts don't.

01:50:17   So you've got to make them come to you with the anger about like, "ATP is using a secret

01:50:23   API."

01:50:24   And then you could say, "Nope, not secret.

01:50:25   It was just so onerous that we didn't think anyone else could do it, but the webpage has

01:50:28   been there forever."

01:50:29   And then they're like, "What can they say then?"

01:50:31   They'd be like, "Oh, well I guess we can do it, but it seems kind of annoying."

01:50:34   And so, you know.

01:50:35   Anyway, I still didn't think you'd get good adoption except in Brooklyn, but that's

01:50:39   something, right?

01:50:40   - Yeah, but they wouldn't even, they wouldn't even want it because like the headphones and

01:50:44   stuff that look really cool, that look cool enough to be in Brooklyn aren't actually

01:50:48   good enough.

01:50:49   Like, you wouldn't even, and the sad part is about all this, the main reason I'd be

01:50:53   doing all this is to make our theme song sound better.

01:50:56   Like our speech, we've already reached the point where our speech is being represented

01:51:00   in a way that is pretty much what I'm putting out from Logic.

01:51:03   You really can't tell the difference between the wave and the MP3 for our speech.

01:51:07   You can only tell for the theme song.

01:51:09   - Can you do multiple enclosures so the player will play the, you know, basically have it

01:51:15   three MP3s, have the show, the song, and then the after show?

01:51:20   - Well, it has to be one file for podcast clients.

01:51:24   You could technically just have basically

01:51:26   like three constant bit rate sections of the file,

01:51:29   but then any seeks during streaming to the after show

01:51:33   would have the wrong time stamps.

01:51:34   And when they're wrong, they're wrong by a lot.

01:51:37   Like in my test of trying to seek an EBR file,

01:51:40   it's off by like a minute and a half.

01:51:41   Like it's a pretty big difference

01:51:44   and has the wrong time stamp.

01:51:45   And then, yeah, it's a mess when it happens.

01:51:48   - So can we go back just a little bit?

01:51:49   So you had said if you created your own version

01:51:53   of this table, you would stuff it in an ID3 tag,

01:51:56   hypothetically?

01:51:57   - Yeah, 'cause it'd most likely be too big

01:51:59   to fit in a frame, and I wouldn't want to run the risk

01:52:02   of a player trying to play the frame as audio

01:52:05   and weird things coming out of the speakers.

01:52:07   So again, it's no big deal to shove it into ID3.

01:52:11   ID3 has a max size of 256 megabytes,

01:52:13   so there's a lot you can shove in there.

01:52:16   - So the hypothetical scenario then would be

01:52:19   you have Marco's custom jump table in an ID3 tag,

01:52:24   but you would still presumably populate

01:52:27   the really crummy existing jump table

01:52:30   that's in that frame.

01:52:32   - Correct, yeah.

01:52:33   - So the, and that would work with no server side changes,

01:52:36   they would just work in Overcast,

01:52:39   and it would fall back and degrade gracefully

01:52:42   in other clients.

01:52:42   But the problem you have with that

01:52:44   is that the Overcast jump to this moment feature,

01:52:49   which hand on heart, no sarcasm intended,

01:52:51   I think might be the most impressive feature of Overcast,

01:52:54   even more so than Smart Speed.

01:52:55   It would break that feature,

01:52:58   and that's why you don't wanna do it.

01:53:00   - Basically, yeah, because I really do think

01:53:03   that my share links are very important for podcasting.

01:53:06   - I agree.

01:53:07   - And not a lot of people use them, but they do get used,

01:53:11   and the usage is going up over time.

01:53:14   That to me is very important,

01:53:15   and I want to keep promoting that,

01:53:16   I wanna keep making them better.

01:53:18   I have a lot of crazy ideas for how to make them better.

01:53:20   Most of these ideas are terrible and will never happen,

01:53:23   or I will attempt them, realize they're terrible,

01:53:26   and then cancel them before I actually release them.

01:53:28   But some of these ideas will actually work

01:53:29   and will be good.

01:53:30   I just don't know which ones yet.

01:53:32   (laughs)

01:53:32   That's how this goes.

01:53:34   I really do care a lot about those times.

01:53:36   And this is one of the reasons why I really get annoyed

01:53:39   with some of the big publishers

01:53:42   using dynamic ad insertion platforms,

01:53:44   because when they do dynamic ad insertion,

01:53:46   which basically gives you new ads on every download.

01:53:49   So if you download a really old episode

01:53:51   of a storytelling show, and you get a brand new ad in it,

01:53:55   and it's like, "Oh wow, this company didn't even exist

01:53:57   "when this episode ran in 2013," or whatever,

01:53:59   that's what's happening is literally every download

01:54:01   they're serving you a new ad, and the idea there

01:54:03   is to better monetize their back catalogs,

01:54:06   because their advertisers paid back in 2013,

01:54:09   they're not getting paid anymore, so they're like,

01:54:10   "Let's put in new ads, we can charge people again."

01:54:13   Fun.

01:54:13   One of the problems with these platforms,

01:54:15   one of the many problems with these platforms

01:54:16   is that they don't always have ads

01:54:19   that are the same length as the original ads.

01:54:20   So basically, timestamps are not persistent

01:54:24   between downloads because the ads you're inserting

01:54:28   in the show are very in length on every download.

01:54:32   So it totally breaks timestamp share lengths,

01:54:34   which drives me crazy.

01:54:36   Like I'm trying to make sharing better

01:54:37   and you're throwing it away.

01:54:39   Everyone complains, like podcasts don't share,

01:54:41   we need more sharing for podcasts.

01:54:43   And then the big podcast producers

01:54:44   make sharing more difficult.

01:54:45   Well, if you really want to solve that problem,

01:54:47   you know, the QuickTime/MPEG Consortium solution

01:54:51   to that problem is that you need to have a more comprehensive map

01:54:55   of the content that also incorporates

01:54:57   the maps of the ads.

01:54:58   So when they change the ads, they change the map.

01:55:00   And so you can do-- it's like source maps for JavaScript

01:55:02   when you minify it, right?

01:55:04   Well, look, you can do that with chapters.

01:55:06   You could store the chapter ID and an offset

01:55:08   within that chapter ID.

01:55:09   Right, but that would mean is the sharing links can't just

01:55:11   contain an offset.

01:55:11   They have to contain an offset in like a version

01:55:13   to say in this version of the file is this offset

01:55:15   and the map has to say, oh, well,

01:55:16   now I've inserted a new ad since then

01:55:17   and I can translate your offset

01:55:19   exactly like source maps on JavaScript.

01:55:20   This offset in this file is actually

01:55:22   this offset in this other file.

01:55:24   - Yeah, no one's gonna do that though.

01:55:25   'Cause this is the thing,

01:55:27   any advancement you make in podcasting,

01:55:30   you have to assume that if it involves any producers

01:55:37   changing their workflow in any way,

01:55:40   or especially changing their CMS in any way,

01:55:43   It's never gonna happen, no one's gonna do it.

01:55:45   - Yeah, you need to have one giant proprietary platform

01:55:47   that can dictate because they're what matter

01:55:51   and whatever they do is what everyone has to follow

01:55:53   and that's how it would work, but we don't have that

01:55:54   and that's a good thing, so you're stuck

01:55:56   in the technological backwater that is the MP3 format, enjoy.

01:56:00   - Honestly, I really do enjoy the format.

01:56:03   The format is very refreshingly simple and straightforward.

01:56:06   It just has this one wart of this stupid 100 byte precision

01:56:12   offset thing for VBR files that, and again,

01:56:15   when they're downloaded, it's not a problem.

01:56:17   It's only a problem when they're streaming

01:56:18   and you're jumping ahead.

01:56:20   - That's a wart on a wart though,

01:56:21   because the original wart is they don't have

01:56:23   this information, and the secondary wart is

01:56:25   we've tried to jam this information in

01:56:27   using tiny little bytes and we can only have 100,

01:56:29   we can only do percentages.

01:56:31   - Yeah, well, the bigger problem though is that

01:56:33   everybody stopped advancing the MP3 format 15 years ago.

01:56:37   - Yeah, it's technological backwater.

01:56:39   (laughing)

01:56:41   - No, and I'm gonna hear from all the AUG

01:56:45   and the HEAAC and the MP3 Pro people.

01:56:49   I'm gonna hear from all these people.

01:56:50   All the newer audio formats.

01:56:52   HEAAC's supposed to be better than MP3.

01:56:55   Yes, I'm aware of all these arguments.

01:56:57   - What about real audio?

01:56:58   Oh, God, you are a bad, bad man.

01:57:01   (laughing)

01:57:02   - I bet they solve this problem.

01:57:04   - No, I mean, the reality is MP3 encoders are so good

01:57:08   and have been so good for quite some time now,

01:57:10   But a well encoded MP3 is already a really fantastic

01:57:15   trade off of size versus quality

01:57:17   and it's compatible with everything everywhere

01:57:21   and it has been for a very long time

01:57:23   and that's why people still use MP3.

01:57:24   Also, there's patent issues with some of the newer ones

01:57:28   and MP3 did have patents on it,

01:57:31   but almost all of the MP3 patents have expired

01:57:34   and the ones that haven't expired, A, expire soon

01:57:37   and B are I think entirely or mostly

01:57:41   for obscure variants that you could enable

01:57:44   that almost nobody does enable of the format.

01:57:47   So MP3 is almost public domain

01:57:50   and it is effectively public domain now.

01:57:52   Because it is so old that it is a technological backwater,

01:57:55   thanks John, it is so old that patents have expired

01:57:58   which makes it way better than everything else in the world.

01:58:01   - Wow.

01:58:02   - You know what's even worse about this too?

01:58:03   What drives me nuts about this?

01:58:05   In my analytics and overcast,

01:58:07   streaming is like 10% of playback.

01:58:09   - Really?

01:58:10   - Like I did all that work for streaming

01:58:12   and now I'm fretting all about streaming.

01:58:14   - I'm glad you did it because now I have overcast

01:58:16   on my iPad and I tell it to stream everything

01:58:18   and download nothing because I don't want it

01:58:19   taking up any space.

01:58:20   - Yeah, actually on an iPad that you listen on sometimes,

01:58:23   that is the perfect case for streaming.

01:58:25   It's perfect, but man, ugh.

01:58:27   It makes me so sad that I did all this work for streaming

01:58:31   and almost nobody uses it basically.

01:58:34   (beeping)