Podcast Search

The Accidental Tech Podcast

205: People Don't Use the Weird Parts

Select a time to play from there. The line being spoken is highlighted as the audio plays.

0:00:00so let's start with some follow-up why

0:00:02did chris lattner decide to leave Apple

0:00:04last week we talked about this and we

0:00:07tried to list all the reasons we could

0:00:09think of why he might be leaving and I

0:00:11think we covered most of the major

0:00:14threads of speculation and didn't really

0:00:15get a listener feedback about anybody

0:00:17did reply was mostly voting for one of

0:00:18the things we talked about but here's

0:00:21the thing i was topic unless there's

0:00:22some kind of like secret doomsday thing

0:00:25about Apple that only Chris knows it's

0:00:27not actually that important that we

0:00:29figure out like why he left apple or

0:00:31what was in his heart of hearts like

0:00:32aside from just being busy bodies are

0:00:34just wanting to know like a gossip type

0:00:35stuff and second pragmatically speaking

0:00:39this is just not the type of thing that

0:00:41people who leave Apple talk about or you

0:00:43know if they do talk about it's like

0:00:45years and years later after everyone no

0:00:46one cares anymore or you know like it's

0:00:48not a big deal

0:00:49so even if there was some deep dark

0:00:51reason it's not like we would know about

0:00:53it until years from now and honestly the

0:00:55straightforward boring explanation is

0:00:56probably the right one anyway so you

0:00:59know I don't know what do you think

0:01:00Chris that sounds about right to me John

0:01:02would you ever buy an electric car in

0:01:07white if you weren't such a jerk

0:01:11why are you so mean to be barca no no

0:01:13I'm fine red cars there you get all

0:01:16yellow thank you get them yellow that's

0:01:18bold

0:01:19my guess alright we obviously need to

0:01:22talk about what's happening right now so

0:01:23ATP is not interested in interview show

0:01:26but in extraordinary circumstances weird

0:01:30things can happen and so in today's

0:01:33episode we have an extraordinary

0:01:34circumstance former apple employee Chris

0:01:37lander was like nice enough to join us

0:01:39on the show and so this is going to be

0:01:42three people interviewing one person

0:01:43which is to say the least a bit peculiar

0:01:46and it's something that we've never

0:01:47really done before so this could be a

0:01:50little bit of a rough ride but we're

0:01:52gonna try to make our best of it and and

0:01:54i'll just say right up front

0:01:56thank you so much Chris for coming on

0:01:57the show and we definitely appreciate it

0:01:59to kind of kick all this off would you

0:02:01mind giving us like a nickel tour of who

0:02:03is crystalline okay well I can try

0:02:07I'm a programmer and

0:02:10of coding i have for a long time I am

0:02:13most widely known for starting lbm and

0:02:16getting it off the ground and handing it

0:02:18off to other smarter people i started

0:02:20playing which is the C C++ objective c

0:02:23compiler and likewise got to the point

0:02:25where it was thriving and then let other

0:02:28smarter people run it and then started

0:02:31Swift and got Swift going and it

0:02:33recently got past three dough and it has

0:02:36an amazing community and has a great

0:02:38team of people running it and now I'm

0:02:40ready to jump to the next thing and so I

0:02:43guess the short version of what i'd like

0:02:45to do is tackle really hard problems and

0:02:48do things that most people think you're

0:02:50impossible

0:02:51so we're gonna go back in time a little

0:02:53bit and i hope that you could tell us

0:02:57how you ended up coming to apple like if

0:02:59you can remember back that far like you

0:03:01know you

0:03:02lvm you started before you or an apple

0:03:04right so how did you end up an apple

0:03:06yeah isn't it a great question so i

0:03:07started working out via aim at the

0:03:09University of Illinois it was a research

0:03:12project and the it really was kind of

0:03:17hope for and designed to be fully

0:03:19general and production quality and all

0:03:21those great things that you want as a

0:03:24graduate student for people to use your

0:03:25code but it also wasn't it was had a lot

0:03:30of the great architectural ideas maybe

0:03:32but it wasn't fully baked it wasn't

0:03:34wouldn't solve all the world's problems

0:03:37and by the time I got to graduate

0:03:41the question was what do I do next and

0:03:43of course my dream was to continue

0:03:44pushing forward and build it out to be

0:03:47something real and to do that i would

0:03:49have to go to a place that would allow

0:03:50me to work on it and I was really

0:03:52fortunate to build relationships with

0:03:54folks at Apple they took a chance on me

0:03:56and I jumped apple and started building

0:03:59out of them is my job was it was totally

0:04:02a dream of mine so that's how you have

0:04:04to graduate

0:04:05did you get your degree and then go to

0:04:06apple yeah yeah that was nice

0:04:09otherwise how do you ever get a job yeah

0:04:11you can get a job that degree we all

0:04:12know that no yeah yeah so

0:04:15how would you explain lvm to somebody

0:04:17who doesn't know stuff about compilers

0:04:19because like we all just rattle off the

0:04:21LEM compiler building tool kit or

0:04:23whatever like but how would you explain

0:04:25that to someone bizarre other listeners

0:04:27of programmers who isn't into may be

0:04:29familiar with how compiler works

0:04:30sure so I mean first we start with what

0:04:32does the compiler a compiler is the

0:04:35thing that takes the code the programmer

0:04:37writes and turns it into something

0:04:39machine can understand and so there's

0:04:41lots of different kinds of computers

0:04:43with lots of different kinds of

0:04:44processors most programmers don't have

0:04:47to think about that or know about that

0:04:49and they wanted program and think in a

0:04:51much higher level than what the the

0:04:53actual processor can do and so the

0:04:55compiler job is to transform what what

0:04:58the human wrote and to something the

0:05:00machine can understand then lvm is a

0:05:04modular system for building different

0:05:06kinds of compilers because there's lots

0:05:08of different reasons why people would

0:05:10want to use compiler technology and I

0:05:12think album was most successful over the

0:05:14years because was built as reusable

0:05:16components kind of like a framework and

0:05:20then the framework can be used to solve

0:05:22lots of different problems and one of

0:05:24the fun things about lvm is that it was

0:05:26eventually over the years used to solve

0:05:28problems that we had never even imagined

0:05:29which is really exciting

0:05:31so one example of that the type of

0:05:33problem is a remember from passwd sees

0:05:36the idea that the xcode ID the GUI that

0:05:40you programs used to to make their

0:05:41applications that very often in that

0:05:44gooey you need to do things that have an

0:05:46understanding of the code but when the

0:05:48compiler is big monolithic thing the

0:05:50best you can kind of do is compile

0:05:52everything in the background and spit

0:05:53out some stuff but when it's a modular

0:05:55toolkit you can take those pieces and

0:05:58just use the part that understands like

0:06:00the syntax tree or whatever and

0:06:01integrate them was that the first use of

0:06:04lvm is something other than like you

0:06:08know a complete like a straight-up

0:06:09compiler that you just point your code

0:06:11and produces output like you know in

0:06:13other words integrating it into

0:06:14applications or was there some use

0:06:15before that

0:06:17well i think the the first use and

0:06:19production was in the opengl stack which

0:06:22end of shipping and a late software

0:06:24update for tiger

0:06:25and as well as in leopard which is 10

0:06:28dot 4 in 10 5

0:06:30there was using a completely invisible

0:06:33way to translate to basically repack

0:06:36data and uploaded to the GPU and so when

0:06:39you need to

0:06:40when you're making opengl calls to say

0:06:42GL vertex 3 f or whatever the the

0:06:45low-level GL call is those verses need

0:06:49to be transformed into a very specific

0:06:51data format the GPU can understand and

0:06:53you have a whole bunch of different

0:06:54kinds of calls that can be used to

0:06:56produce for tech surge in other geometry

0:06:58data but you also have a matrix of

0:07:00different kinds of GPUs you need to

0:07:02support and they all have different

0:07:03formats and different capabilities and

0:07:05requirements and so Alabama's used to

0:07:07generate very small chunks of code to do

0:07:09that and it was part of the 64-bit bring

0:07:12up in the leopard timeframe and so that

0:07:14was probably the first completely

0:07:16invisible use of it the that used and

0:07:19proved a lot of the kind of code

0:07:21generation technologies so forgive me

0:07:24one more lvm related things are going to

0:07:26go from what is the compiler all I down

0:07:27to the nitty-gritty stuff here my vague

0:07:30recollection of what you just described

0:07:31the whole video driver thing was that

0:07:33there were files on disk in the shipping

0:07:36us that were basically lvm bytecode like

0:07:39these little bc files and they would get

0:07:42slurped up and converted to machine code

0:07:45appropriate for the GPU and spaz is that

0:07:48am i right about that yeah you're

0:07:49absolutely right and so it basically the

0:07:51we look at is the the code files the lvm

0:07:55the ship with the OS were little

0:07:57snippets of code that then at runtime or

0:08:00recombined and optimized across and so

0:08:03that was kind of the library of

0:08:04primitives that the opengl runtime used

0:08:07to assemble what it was trying to do

0:08:09alright so my question about this and

0:08:11maybe you kind of alluded to with either

0:08:12a slip of the tongue or reviewing

0:08:14understanding is what is the difference

0:08:16between llvm bytecode and what Apple

0:08:18currently brands as bit code

0:08:21oh well DRS first or quote is a

0:08:23historical it's a technical and

0:08:26historical cuz I'm confused by the

0:08:28branding and I'm also confused by the

0:08:29tech so lvm prior to lv m2 dot oh and

0:08:35these version numbers of the open-source

0:08:36version numbers not related to the apple

0:08:38version

0:08:38numbers and their confusingly different

0:08:40for not very interesting reasons but

0:08:43before to do used a really horrible

0:08:46encoding in binary the was branded

0:08:50bytecode because that's what java use

0:08:52had a whole bunch of problems and there

0:08:54was an extensible it was fragile with

0:08:57the ilbm was changing and didn't have

0:09:00compatibility and so in the lvm to do

0:09:04timeframe which josh is probably don't

0:09:08know ten ten years ago or more it's been

0:09:10a long time ago now I redesign the whole

0:09:15thing and it now uses this quote-unquote

0:09:17bit code format and use bit code as a

0:09:20term for two reasons one of which was it

0:09:22was different than bytecode and so the

0:09:24two could be side-by-side in the

0:09:25directory tree for a while and the

0:09:27second reason is that the encoding is a

0:09:29bitstream format instead of a byte

0:09:30stream but that that's the origin of the

0:09:34term bit code you can find documentation

0:09:36in this file format is actually a very

0:09:38general encoding format that you can use

0:09:41for xmltype purposes it's a

0:09:43self-describing file format and multiple

0:09:46different things can be encoded in a bit

0:09:48code that was definitely some confusion

0:09:50about that because when Apple presented

0:09:51it when they presented it as like the

0:09:53required was like the watch apps have to

0:09:54be a bit code Marco you would know this

0:09:56watch's required iOS is still optional

0:09:59and TV is required and they would have

0:10:02you send you put your stuff in bit code

0:10:04and there's a lot of speculation early

0:10:06on about what that meant like in terms

0:10:09of does that mean that what you're

0:10:12uploading is not a finished binary and

0:10:14therefore it's like this machine

0:10:15agnostic and could be targeted any CPU

0:10:17and the more prosaic version like it

0:10:21doesn't mean its CPU agnostic all it

0:10:23means is that Apple has slightly more

0:10:25freedom to like change instructions on

0:10:28the sea views they do target

0:10:30how would you characterize the what what

0:10:32what advantages are there of compiling

0:10:36something to bit code and then uploading

0:10:38it somewhere vs sending someone a binary

0:10:40sure there's a good there's a couple of

0:10:41different advantages for that one of

0:10:43which is that the compiler keeps getting

0:10:46better and so the compiler learns a new

0:10:47optimization at

0:10:49can be applied to existing applications

0:10:51without application developers and

0:10:52reupload their their app that's one

0:10:55small thing other things is our the

0:10:58apple just periodically add new

0:10:59instructions to its at cps one example

0:11:02this historically was the the

0:11:06hilariously named swift check that

0:11:08launched which was the first designed

0:11:11in-house 32-bit ARM chip this was the

0:11:15iphone 5 if I recall that's right

0:11:17so in this chip they added an integer

0:11:20divided instruction and all the chips

0:11:22before that didn't have the ability to

0:11:24do an integer divided in hardware you

0:11:26had to actually open code it and there's

0:11:28a library function to do that and so

0:11:30that and the other instructions they

0:11:32added were a pretty big deal and use

0:11:35pervasively and the only way to tackle

0:11:37then in and handle that and use it

0:11:39effectively was to introduce an entire

0:11:41new slice and so if you're an app

0:11:43developer back then you had the v7 slice

0:11:45which was all the older ships and then

0:11:48you have to be 7s slice as for swift

0:11:50which was the slice that included the

0:11:55injured but now this is a pain for app

0:11:57developers is paying for Apple everybody

0:12:00has it has to convince all the

0:12:01developers to take it seriously and if

0:12:03it doesn't if app developers don't

0:12:04support it then the new hardware doesn't

0:12:05benefit from it and so--but code allows

0:12:08that kind of thing to go completely away

0:12:10with the code

0:12:11apple just recompile the applications in

0:12:13the store and say haha this this device

0:12:16has an integer divided instruction and

0:12:18it can it works perfectly with the

0:12:20thinning kind of architecture they have

0:12:21for deploying resources to specific

0:12:24device types and all kind of looks

0:12:26together bit code is not a magic

0:12:30solution though it doesn't you can take

0:12:32a 32-bit app for example and run on a

0:12:3564-bit device that that kind of

0:12:37portability isn't something that can

0:12:38give you

0:12:39notably because that is something that's

0:12:41visible and see and so as you're writing

0:12:44C code you can write if death you know

0:12:47pointers equals 32 and that's something

0:12:49that big coke and abstract / and so it's

0:12:53useful for very specific low-level kinds

0:12:55of enhancements

0:12:57but isn't a panacea that makes

0:12:59everything magically portable at the

0:13:02same thing i would assume for like

0:13:03architecture changes especially there

0:13:04was an Indian difference because Indian

0:13:05this is visible from the sea world so if

0:13:07you can't you can't target different

0:13:08ending this

0:13:09yep yeah so this is not something

0:13:11magically Saul portability problems but

0:13:13is it is very useful for specific

0:13:15problems that Apple's face in the past

0:13:16and then going back to the this life

0:13:19thing that's like terminology like I'm

0:13:20tamaco format where you get to put

0:13:22different architectures worth of

0:13:23executable or is that a different

0:13:25yeah it is i think this technology that

0:13:26goes all the way back to next when next

0:13:29supported really heterogeneous different

0:13:31architectures like pa-risc and and spark

0:13:33and RBC and Intel all back and in the

0:13:37next days and it allows you to compile

0:13:40your application with multiple slices

0:13:42they call them and then there's a tool

0:13:44called lipo which sticks them together

0:13:46and at runtime the OS pics the slice the

0:13:51best matches your cpu and practice and

0:13:53so an app developer can support many

0:13:55different kinds of architectures and

0:13:57then the OS just does the right thing

0:13:59and so in today's world it's very common

0:14:01to build an iOS app for example the for

0:14:05both arms 64 and arm 32 and the way that

0:14:07works in the native in the simplest

0:14:10cases it's just too different binary is

0:14:13that are stuck together and then the OS

0:14:15X the right one to run all right i could

0:14:18go on for this river but I want to

0:14:20that's right all it's just uh is this

0:14:21some pent-up relations about Bitcoin

0:14:24biko but now I feel satisfied we can go

0:14:26on to broader topics we sponsor tonight

0:14:30by Squarespace squarespace.com use code

0:14:34ATP to get ten percent off your first

0:14:36purchase

0:14:36it's the new year you might have a new

0:14:38project you might have a new goal you

0:14:40want to reach and that might involve

0:14:41making a website make your next move

0:14:44with a beautiful website from

0:14:46Squarespace Squarespace sites are so

0:14:49incredibly easy to make and they're

0:14:50incredibly beautiful and the tools are

0:14:53very powerful

0:14:54this is a great combination that i

0:14:55personally have never seen before in the

0:14:58web building space until Squarespace the

0:15:00amount of power you get with these tools

0:15:02is just incredible you can make so many

0:15:04different kinds of sites with so little

0:15:06effort

0:15:07nothing else comes close your site's

0:15:10look professionally designed

0:15:11regardless of your skill level with no

0:15:13coding required although if you want to

0:15:15jump in and help some of the code you

0:15:16actually can but you don't have to and

0:15:18it probably isn't a good use of your

0:15:19time because Squarespace is such a great

0:15:22platform right out-of-the-box whether

0:15:25you're making website for you or for

0:15:26somebody else if you make of somebody

0:15:28else

0:15:29Squarespace is the place you want to do

0:15:30that because you shouldn't have to use

0:15:33Squarespace and then you're done

0:15:35they're not coming to you anymore

0:15:35squares page will support if they need

0:15:37help and able to figure things out on

0:15:39their own because it's just so much

0:15:40easier to use than whatever CMS you were

0:15:42going to install on their server so

0:15:43check out today go to squarespace.com

0:15:45use code ATP to ten percent off your

0:15:48first purchase when you decide to sign

0:15:50up whether you're making a site for you

0:15:52or somebody else make your next move

0:15:54with squarespace ah so Chris how's your

0:16:00transition from coding to being a

0:16:03manager because from an outsider's point

0:16:04of view it looked like you know you

0:16:06arrived at Apple you were working

0:16:07heavily on llvm like you said earlier

0:16:09but over time I mean even just looking

0:16:11at the resume on your website it seems

0:16:13pretty clear that you are doing a lot

0:16:15more managing lately then perhaps coding

0:16:18or maybe maybe just worked around the

0:16:20clock that would surprise me either but

0:16:21it how how would you how did you feel

0:16:24about your transition from from being

0:16:26just a code monkey to being a manager

0:16:28and I asked in part because at my face

0:16:31my career I am still a code monkey but i

0:16:32think it's not too far away that I might

0:16:34become a manager and so I'm curious to

0:16:37hear you know from from one code monkey

0:16:39to another how did that transition go

0:16:41and what did you learn from it

0:16:43yeah that's that's kind of interesting

0:16:44so I don't think it is a hard transition

0:16:47I'm still writing code i still love

0:16:49writing code that's something that I

0:16:51think is a key part of my identity and

0:16:54but on the other hand is it is very

0:16:58strange so I started managing people at

0:17:00Apple really pretty early on maybe

0:17:03you're so after i joined i started

0:17:06managing two or three people because

0:17:09kind of the best person to to manage the

0:17:12nascent lvm effort at the time and

0:17:16initially it was mostly just being a

0:17:19tech lead and

0:17:20so if you're if you've ever driven a

0:17:22project where you have co-workers that

0:17:25don't report to you that your kind of

0:17:27guiding them and helping make technical

0:17:29decisions that's that's where started

0:17:31over over time I my team grew and

0:17:35eventually ended up picking up and

0:17:37running kind of a second level manager

0:17:39role where you're managing other

0:17:41managers and each step along the way it

0:17:43introduces new challenges and along the

0:17:46way I told myself that managing

0:17:49something that I had to do because that

0:17:52was the best qualified to do it and I

0:17:54didn't want anybody else to and that

0:17:57coding was my real joy and passion and I

0:17:59think the coding really is my joy and

0:18:01passion but along the way

0:18:03looking back I realize that I am

0:18:04actually pretty good at managing and I

0:18:07really do like building teams i like

0:18:10working with people and I've never

0:18:11considered myself to be a people person

0:18:13just how it's just very very odd for me

0:18:16to even think about that but actually do

0:18:19understand how people work and can you

0:18:22know really care about getting the most

0:18:24out of people and working with them to

0:18:26make them successful and so it's been

0:18:29kind of a transition over time but I've

0:18:32always hung onto coding and even though

0:18:34i have a full time management job I

0:18:36could like crazy and nights and weekends

0:18:38and I'm not the kind of guy that likes

0:18:40around idle and eaten and so I've always

0:18:43just been busy so I mean you can get as

0:18:46much of this history as you want to

0:18:47Chris but how did you get Swift to

0:18:50happen inside Apple because from the

0:18:53outside its mysterious to us how big

0:18:57important projects come to be like you

0:19:00know whatever you name the headlining

0:19:02things that are common Apple we hear

0:19:04that Apple does a lot of things that

0:19:05don't see the light of day

0:19:07how did you get probably one of the

0:19:09biggest things ever to come out of Apple

0:19:11an entirely new language created

0:19:13in-house had to get that to happen and

0:19:14not just be some little thing that was

0:19:16off to the side that fizzled often

0:19:18disappear

0:19:18sure well so i can tell you about Swift

0:19:21but i don't think you should project

0:19:23this onto every other project and apple

0:19:24because i'm sure there are different so

0:19:26you just talk about my experiences so

0:19:29started in 2010

0:19:31you know it's kind of the timing is

0:19:35suspicious because right after a blogger

0:19:37wrote something about how Apple need a

0:19:40new programming language haha i'm not a

0:19:43blogger ok

0:19:44arstechnica wasn't a thing about blog

0:19:48that was a prestigious website still is

0:19:50ok ok hi I stand corrected

0:19:54so I don't remember fight after i read

0:19:58it exactly then that's that was a causal

0:20:00link to starting working on swift or if

0:20:02it was just because I just finish up

0:20:05cleaning c++ support and we launched

0:20:07client c++ and of course if you write

0:20:10C++ good you know how ugly and horrible

0:20:12it is but if you implement c++ it's a

0:20:15whole nother level of the of challenges

0:20:17and so getting that first release out

0:20:20was it was a major accomplishment for

0:20:22for the entire team and it was a major

0:20:23major effort for a ton of people

0:20:26involved and as is typical you get

0:20:29something over the finish line and of

0:20:31course it wasn't fully finished long

0:20:32ways to go but a lot of the

0:20:33intellectually interesting to me at

0:20:36least pieces were we're done

0:20:38that's why I started thinking about okay

0:20:40what comes next

0:20:41right and then implementing other

0:20:43people's stuff like for example the see

0:20:46the c++ the objective-c standards and

0:20:49languages but there were there were you

0:20:53know things that we weren't happy with

0:20:54and things I wasn't happy with and so

0:20:56what started out as just a let's let's

0:20:59see what could be done right let's see

0:21:01what a much simpler language that

0:21:03doesn't need a preprocessor doesn't hate

0:21:05try graphs doesn't need all the weird

0:21:08things that have accumulated into c and

0:21:10c++ over the years but let's see what we

0:21:13can do and just try to build something

0:21:15nicer

0:21:16initially it was really just a me me

0:21:20messing around and nobody knew about it

0:21:23because it wasn't anything to know about

0:21:25but eventually got a little bit more

0:21:28serious and said hey well after playing

0:21:30around a little bit i think that this

0:21:32actually could make sense and so I

0:21:34started talking to my management

0:21:36I'm start talking again engineers are

0:21:39working on playing and they seemed

0:21:41excited about it we got a couple people

0:21:43working on it part time and I can ask my

0:21:46manager that it was interesting enough

0:21:47that we could have a couple of people

0:21:50work on it now this wasn't a major

0:21:53commitment this was a ok well let's see

0:21:56see what there is here and that started

0:21:59a an internal demo schedule and other

0:22:02things we set milestones in place and

0:22:04tried to to to justify the investment

0:22:06that was being put into it eventually

0:22:09got to the point where there was an in

0:22:12it was part of a long that it was posing

0:22:14strategic questions so the strategic

0:22:17questions were the form

0:22:18ok what we have objective-c objective-c

0:22:21is a great language it is probably

0:22:24directly responsible for the iphone

0:22:27being successful because it's really

0:22:31high performance allows you to get your

0:22:33job done it supports high-level

0:22:35frameworks and beautiful way there's a

0:22:37ton of stuff to love about objective-c

0:22:39and while there are few things that are

0:22:41ugly about it some at sides and

0:22:42semicolons and other stuff like that you

0:22:44know we can make objectively better and

0:22:46so the question was always why not just

0:22:49make objectively better right why don't

0:22:51we just keep evolving objective-c and

0:22:54why do we want to face potential

0:22:55disruption in terms of moving the entire

0:22:57development community to something new

0:23:00and we kick that around for a long time

0:23:02we talked about both sides and we came

0:23:05to realize that yes we can and should

0:23:06make objective-c better and we continue

0:23:09to invest in objective-c we did things

0:23:11like our for example which is a major

0:23:13effort but we got dots

0:23:17well so dot syntax / dated this ah yeah

0:23:20dot dot syntax so the objective c 2

0:23:23features and the block syntax were all

0:23:25kind of prior to this i worked on blocks

0:23:28but

0:23:28but it wasn't in this time for him and

0:23:32so we were talking about okay well can

0:23:34we just make objectively better and can

0:23:35feature creep into the language that we

0:23:37want for the fullness of time because we

0:23:39can that be much less disruptive to the

0:23:41community and we decided that yeah we

0:23:43can we can move objective-c a lot closer

0:23:45to what we want so we can get automatic

0:23:47memory management with art for example

0:23:49but we can't ever take away the problems

0:23:52that lead to objective-c being unsafe

0:23:55and the fundamental problem was

0:23:57objectives he was built on top of C and

0:24:00C inherently has pointers it has

0:24:02initialize variables that has array of

0:24:05overflow that has all these problems

0:24:06that even if you have full control of

0:24:09your compiler and tools back you just

0:24:11can't fix right to fix to fix dangling

0:24:14pointers you'd have to fix lifetime

0:24:15issues and she doesn't have a framework

0:24:17to to reason about that and retrofitting

0:24:20that into a compatible way into the

0:24:21system just one really work if you took

0:24:24away see for objective-c you can't use

0:24:26see arrays on the stack for example and

0:24:28if you can do that there's entire

0:24:31classes of applications that where the

0:24:33performance just wouldn't be acceptable

0:24:34and so we went around around around we

0:24:37said okay

0:24:38the only way that this can make sense in

0:24:42terms of the cost of the destruction to

0:24:44the community is if we make it a safe

0:24:47programming language which means not

0:24:49safe as in you can have no bugs but safe

0:24:51in terms of memory safety while also

0:24:54providing high performance and and

0:24:56moving the programming model forward and

0:24:58so that was really kind of the ideas

0:25:00that that came together to make swift

0:25:03being worth the investment and being

0:25:05worth being disruptive to the community

0:25:08so a lot of these kinds of pitches and

0:25:10ideas were being held and very small

0:25:14small small meetings and coming out of

0:25:19WBC 2013 is one we in the executive team

0:25:23decided okay let's let's really commit

0:25:25to this and that's coming out of WC 2013

0:25:28is when the developer tools team came to

0:25:31know about it and really started working

0:25:33hard on it and Swift is Swift as a huge

0:25:36huge effort by hundreds of people

0:25:39and that's when all the work on bringing

0:25:41up the debugger in the IDE support and

0:25:44playgrounds and all the other things

0:25:45that the charge into swift one auto

0:25:47really started coming together and and

0:25:50make happen so it's kind of house which

0:25:53came to be lose again one step after

0:25:55another building small things explaining

0:25:58to people why make sense why was worth

0:26:00it and then also I think that a major

0:26:02important aspect of Swift being I guess

0:26:07worth taking a risk on was that we told

0:26:11people that we would not guarantee

0:26:12source compatibility in window and that

0:26:16was really interesting because that

0:26:19actually was a big relief for people

0:26:20because internally that meant that it

0:26:24didn't have to be perfect when it

0:26:26launched it meant that we could learn we

0:26:27could adapt and that as long as we're

0:26:30clear with the community about that that

0:26:31that would allow us to get to ultimate

0:26:34greatness and learn from actual usage

0:26:37and practice i think that was a really

0:26:39key piece to Swift coming in actually

0:26:42making it to the market so strategically

0:26:45speaking when you were doing like the

0:26:48work to improve objective-c did you have

0:26:51in mind the ability to potentially

0:26:53leverage some of that work for swift

0:26:56like we're like I'm trying to like the

0:26:58strategy of how you get it done like

0:26:59part of his personal relationships and

0:27:01like knowing the right people

0:27:03part of it is persuasion you have to

0:27:04persuade them with your technical

0:27:05arguments in your business arguments and

0:27:06stuff like that but part of it was part

0:27:09of it also laying the groundwork with

0:27:12with efforts and technologies that have

0:27:15multiple uses

0:27:16yeah absolutely so so long for example

0:27:18that is our it was really clear to me

0:27:21that if if we get to memory safety we

0:27:23had to have automatic memory management

0:27:24and I don't know if you wanna go down

0:27:27the GC vs arg rabbit hole or not that

0:27:29part

0:27:30arc seem like obviously the right models

0:27:32me and to enter some other people and so

0:27:35we said okay we need to get to an

0:27:38automatic memory management model

0:27:41objective-c isn't because objective-c at

0:27:43the time had the are obviously garbage

0:27:45collector which had numerous problems

0:27:47and so we had to get objective-c to be

0:27:51memory manage and so that's why ark was

0:27:53who became a thing it became really high

0:27:56priority and I think that it's a great

0:27:59example of something with by itself had

0:28:01a huge amount of value to the

0:28:02objective-c community and then after

0:28:05that we had similar other efforts like

0:28:07modules for example that was a huge

0:28:09thing that both improved build time but

0:28:12was also absolutely essential to

0:28:14enabling swift code to eventually just

0:28:16say import UI kit and get everything and

0:28:19so there's a number of things that

0:28:20rolled out over the years and it was

0:28:22really funny at the time because the

0:28:24objective developers externally and even

0:28:27internally to apple always accused the

0:28:29compiler and languages team of taking

0:28:31objective-c on this random walk and they

0:28:33didn't know how to get together and

0:28:34we're where's the strategy and why are

0:28:37you guys working on this and why are you

0:28:38working on some other syntactic sugar

0:28:40for the thing i want and of course we

0:28:42could never tell them but there was

0:28:44there's a good reason and so on another

0:28:46example that is the object literals

0:28:49feature i think we introduce that in

0:28:502013 fire call work for Ray literals and

0:28:55other things like that that was

0:28:58interesting feature in that it was a

0:29:00relatively small amount of engineering

0:29:03work the broad objectives the objective

0:29:06language much closer to Swift in terms

0:29:09of having arrays and dictionaries

0:29:11literals as part of the language but the

0:29:14real reason for doing it is that it was

0:29:16a small amount of work that move

0:29:18directly forward which allowed the

0:29:19compiler language team to focus on swift

0:29:21because with was a very large amount of

0:29:25work and that was you know at appeasing

0:29:28community and your route solving real

0:29:31problems but you know buying time to

0:29:34actually pour into the multi-year

0:29:35project of making this with language

0:29:36happen

0:29:37that's when you pull out the syntactic

0:29:39sugar programmers love it takes

0:29:40attendance implement here you go that

0:29:43everyone applauded when they see the

0:29:44slide ba I mean in all fairness that was

0:29:47a really big deal anyway so it's kind of

0:29:49building on on the Apple internal

0:29:51reaction thing

0:29:53you know with this this language in its

0:29:56development and then its public release

0:29:58is probably fairly unique language in

0:30:01like the context in which was released

0:30:02you have this this massive company

0:30:05nobody probably had more absurd the

0:30:07objective-c programmers an apple does

0:30:08and see this massive company with tons

0:30:11of internal programmers in a massive

0:30:13internal code base that is using this

0:30:15this other language you have this entire

0:30:17community of iOS and mac developers and

0:30:21then you have the entire world watching

0:30:24every move apple makes and putting under

0:30:26very very heavy scrutiny so the release

0:30:29of his language in this context is

0:30:30probably only very high pressure but

0:30:32also like unique in that you you knew

0:30:36that you were gonna face a lot of people

0:30:39up front like most languages probably

0:30:41start out with a few people using them

0:30:42you know something some random quarter

0:30:44the internet and eventually maybe

0:30:45someone it maybe it slowly takes off

0:30:46this was not that way and so what

0:30:50considerations I guess when into that

0:30:52and also how were you able to convince

0:30:55people even inside of Apple to to use

0:30:59swift and and how is how is that going

0:31:03so I guess there's different answers

0:31:07that four different phases we had no

0:31:12idea that the reaction would be as

0:31:14strong and as positive as it was and so

0:31:16I think we were blown away and caught

0:31:18unaware in some ways but otherwise I

0:31:24think we did the right thing so for

0:31:26example if you look at swift overtime so

0:31:29if one and so if two are as similar to

0:31:32objective sees we can make them so if

0:31:35you look at the way that selectors are

0:31:36named for example it was very there's a

0:31:40very direct correlation if you look at

0:31:42the way objective-c swift one was

0:31:45explained to people it was explained to

0:31:47people as it's just like objective c you

0:31:51can use all your same patterns you can

0:31:52use all same framework she can continue

0:31:54doing object-oriented programming you

0:31:56can do all these things but it also has

0:31:58some cool new things like enums and

0:32:01pattern matching and

0:32:02these other things and so it's great

0:32:04because it's just cleaned up syntax for

0:32:06what you already know

0:32:07plus it has some new cool stuff that you

0:32:09can play with right and then you fast

0:32:11forward a year or two and the Swift

0:32:13community was really taking off

0:32:15it was people were understanding what's

0:32:17with those about Swift open sources also

0:32:20enabled this and so 3 is really about so

0:32:23it's becoming its own language that was

0:32:25really true to itself and standing you

0:32:28know stands by itself and that's where

0:32:30you see the method call syntax for

0:32:32example being radically changed and a

0:32:35lot of things just kind of coming

0:32:36together and really focusing and

0:32:40polishing the language in terms of the

0:32:43the rise of the community the the

0:32:45external community and people using it

0:32:47that was really really hard for us

0:32:50because so my experience has been with

0:32:53lv and with clang clang as a C and C++

0:32:56compiler and objective-c was completely

0:33:00different than Swift because when we

0:33:02open source and launch claim which was

0:33:05at Apple everybody yond open source it

0:33:12very early on and people said okay

0:33:13you're nuts nobody can build a c++

0:33:15compiler that's that's not a thing why

0:33:18are you even messing around with this

0:33:19and you know we got one contributor to

0:33:22the project and we had no users and then

0:33:25we got you know month later we might get

0:33:26another contributor or another person

0:33:28part-time the setting and patches and it

0:33:31was a very slow growth that that was

0:33:34ultimately really great because it was

0:33:37relatively low pressure was swift and

0:33:39with the the way was launched it was

0:33:41completely different because it went

0:33:43from nobody knowing about to everybody

0:33:45knowing about it overnight literally and

0:33:48there being so much excitement so many

0:33:51people that were nervous and had bad

0:33:53reactions other people that were super

0:33:55excited but a lot of people that were

0:33:57suddenly hitting bugs and having

0:33:59challenges and problems and why did you

0:34:01do this and why didn't you do that and

0:34:03you know some of that was misguided just

0:34:05because they didn't know Swift yet but

0:34:07other others other aspects were really

0:34:09spot-on and that was 12 really started

0:34:12learning about how people are using and

0:34:14started

0:34:14during moving Swift and Swift between

0:34:17the WBC lon chance with one auto release

0:34:21change dramatically and that was a

0:34:24really really hard time for the entire

0:34:26team because we really want to get it as

0:34:29close as we could what we wanted but we

0:34:31had a very short amount of time to do so

0:34:33and I think that's been true first with

0:34:36two and so with three as well where we

0:34:38have really high goals for the team and

0:34:41really high goals for what we want to

0:34:43get done and that makes it a lot a lot

0:34:45of pressure and a lot of stress but it's

0:34:49I think worth it to serve and see the

0:34:52community around Swift grow and see

0:34:55people use it and see people learn it

0:34:57and it's been a lot of fun so I'm

0:34:59curious

0:35:00I know when Swift first came out it was

0:35:02hard to get a lot of adoption inside of

0:35:05apple for various you know tooling

0:35:07reasons uh among men maybe whatever else

0:35:09was going on

0:35:10how is internal adoption going today

0:35:13like re seeing meaningful adoption hours

0:35:15at most i was programs were using it and

0:35:17not anybody inside the White team itself

0:35:20has specific goals they need to achieve

0:35:23before there can be truly

0:35:24across-the-board adoption at Apple's ABI

0:35:28stability is the number one thing that

0:35:30prevents framework developers for

0:35:31example from adopting Swift so that's a

0:35:34really important thing that's one of the

0:35:36reasons it's always a really high

0:35:37priority but switch has been adopted by

0:35:40application developers and other things

0:35:42the doc is publix with playgrounds app

0:35:45is public the music app in iOS is

0:35:48publicly known so they're they're

0:35:49definitely some big big adopters more

0:35:52broadly though that the big problem is

0:35:53that i think and i will speak for

0:35:56everybody but in many many people doing

0:35:59objects development and Apple are

0:36:01chomping at the bit they want to be

0:36:02using Swift and so it's it's really

0:36:04matter just getting the technology

0:36:06problem-solve didn't get you know

0:36:07checking out the things are holding

0:36:09people back it's not about you know

0:36:12people dragging their feet and not

0:36:14wanting to use it makes sense that's

0:36:16really also get it from from the outside

0:36:18it's so hard to figure out you know is

0:36:20is Apple internally like oh really

0:36:23or for example internally like you said

0:36:25like really excited and I'm sure that

0:36:26there's some of both but

0:36:27but that's really good to hear yeah and

0:36:29i think it's also changed over time when

0:36:32it first came out of course it was a

0:36:34huge surprise for most of the software

0:36:36team and so you know many people went

0:36:39through the usual phases of matter and

0:36:42no objective c why would have learned

0:36:44this new thing objective-c is great and

0:36:45objective-c is great so that that's

0:36:48that's an objective statement right but

0:36:52I think things have shifted over time

0:36:53and as switches matured it's definitely

0:36:56become a lot more appealing and one

0:36:58other technology problem that is

0:37:00hilarious but also really important is

0:37:03the Apple frameworks that has to support

0:37:0632 the mac apps in 37 mac apps have this

0:37:09interesting challenge of they have

0:37:11what's called the classic objective-c

0:37:13runtime which doesn't support things

0:37:16like non fragile instance variables and

0:37:18things like that and so at some point

0:37:20I'm the Swift team will need to make the

0:37:22Swift runtime work in that mode or

0:37:24figure out some other solution to adapt

0:37:27it because until that happens it won't

0:37:29be possible use Swift an appt get for

0:37:31example

0:37:32well that's fascinating so kind of

0:37:35taking a more broad view do you could

0:37:40you name either you're most proud

0:37:42accomplishment during your time at Apple

0:37:44that you can discuss publicly anyway or

0:37:46if not you're most proud just won the

0:37:48first one that jumps to mind

0:37:49wow I don't have been really fortunate

0:37:53to work on so many really important

0:37:55projects and I mean I think that between

0:37:59lvm and swift for example it's really a

0:38:03hard call

0:38:03I'm Swift I think impacts more people in

0:38:08terms of the number of people would

0:38:10actually know they're using it and want

0:38:13to learn it and things like that but i

0:38:14think the lvm has been picked up and use

0:38:17so far across industry and its kind of

0:38:19standardizing the world's compilers I

0:38:21think that's probably more profound at

0:38:23this point so it's really it's a start

0:38:27calling and that they're there are a ton

0:38:29of really really great opportunities

0:38:33that I've had and spend a lot of fun

0:38:35over many years now

0:38:36also outside the open-source pieces

0:38:38working with the expert team and working

0:38:41with the other teams and developer tools

0:38:42has been really phenomenal

0:38:43I mean there's so many great people at

0:38:45Apple and Android developer tools it's

0:38:47just been a joy

0:38:48so what do you expect to miss the most

0:38:51about being an apple and I'm gonna take

0:38:53I'm gonna let you expand on this but I'm

0:38:55gonna take it off the table and so you

0:38:56have to pick like your second choice

0:38:57you're gonna say like you miss the

0:38:58people because of course you do right

0:39:00yes that is actually what i would say

0:39:02right but everybody you know so you can

0:39:04go ahead and tell us about all the great

0:39:05people your example but besides that

0:39:07what do you expect to miss most about

0:39:09Apple the company working well so the

0:39:11great thing about Apple and the thing

0:39:13that I don't think that many people

0:39:14really get about Apple is it is really

0:39:19one of the few best place in the world

0:39:22where you can work hard but also work on

0:39:25products that ship and the ship to

0:39:29millions of people so that your hard

0:39:31work ends up mattering right there

0:39:33there's a lot of places you can work

0:39:35hard there's a lot of places that you

0:39:37can work on interesting things but

0:39:40there's also a lot of places end up not

0:39:42shipping a lot or or if they ship it

0:39:45very few people actually use it and so

0:39:47apples really really magical in that

0:39:50combines being you know being able to

0:39:53work hard and in do great things with it

0:39:55actually mattering and to me that that's

0:39:57the thing that that is just so so

0:39:59phenomenal about Apple so so moving on

0:40:03to your new role at Tesla do you think

0:40:06that Swift will still be a major part of

0:40:09your work life in some way like you know

0:40:11or will basically be like something you

0:40:13do in the evenings and weekends because

0:40:15we all know that that Silicon Valley

0:40:16tech startups don't usually work people

0:40:18too hard you're gonna have tons of free

0:40:19time

0:40:20yeah so this jump on making is

0:40:22definitely do a cushy no easy that right

0:40:26that that's what I'm looking for here uh

0:40:30so i don't know we'll have to see i

0:40:33haven't started yet and i think it's

0:40:36going to be very challenging job and one

0:40:37of the reasons I'm excited about it

0:40:39because I really want to throw myself

0:40:40into another heart technology problem

0:40:43and so I'm really excited about that i

0:40:45do expect to be involved with the design

0:40:47of swift and helping helping with the

0:40:50core team and helping Swift evolution

0:40:51process and helping in and allow that

0:40:53kind of respect but one of things to

0:40:55keep in mind is that Swift was actually

0:40:57really tiny slice of my job at apple and

0:41:01so you know I had a very full schedule

0:41:05dealing with lots of other stuff and it

0:41:08was already kind of nights and weekends

0:41:09project that i was coating on it and

0:41:12contributing at the level that I was so

0:41:15I'm hoping with all be a big change but

0:41:17honestly i'll have to see obviously I

0:41:20care a lot about Swift and I really want

0:41:22to get to its go over world domination

0:41:25so we're sponsor tonight by eero go to

0:41:28eurocom at the ER o.com and use code ATP

0:41:32when you order for free expedited

0:41:33shipping Wi-Fi is broken

0:41:36you probably have one Wi-Fi router with

0:41:38if you're lucky a bunch of antennas on

0:41:40that look hideous and you're trying to

0:41:43cover your whole house with Wi-Fi but it

0:41:44just doesn't reach every room

0:41:46it just doesn't reach everywhere because

0:41:47Wi-Fi is hard when you're recovering

0:41:49from 1.0 solve this problem by having a

0:41:53router system where they will they had

0:41:55this little ear units nice attractive

0:41:56it's tastefully designed there's not a

0:41:58bunch of antennae sticking out of it and

0:42:00you can buy one of them or you can buy

0:42:0210 of them are a number in between and

0:42:04that you put them around your house and

0:42:06each one covers the area in Wi-Fi and

0:42:09they all talk to each other

0:42:11wirelessly so you don't have to run

0:42:12wires love your house you can have your

0:42:14entire house blanketed in Wi-Fi with 0

0:42:16and they talk to each other over this

0:42:18mesh network so it's even faster than

0:42:20previous if you ever use like an

0:42:21extender and centers are very slow

0:42:23it's even faster than that by a lot and

0:42:260 is true enterprise-grade quality and

0:42:28performance they have everything you

0:42:30need from encryption features two things

0:42:32like parental controls and that they're

0:42:33actually updating it over time they've

0:42:35done over 12 updates since launch with

0:42:37more to come to add features they have

0:42:39an app for very easy setup and they have

0:42:41incredible customer support if you need

0:42:43it

0:42:44they recommend 10 per thousand square

0:42:45feet so typical home will have two or

0:42:47three of them they got a three pack as a

0:42:49pretty reasonable starting point for

0:42:50most everybody there's a 30 day

0:42:52money-back guarantee if you don't like

0:42:54it or if you could buy too many you want

0:42:55to return a couple of them if you

0:42:56I'm not eating as many as you bought

0:42:58check it out the reviews are stellar

0:43:00they back this up go to 0 comments de ro

0:43:04dah calm and use code ATP when you order

0:43:07for free expedited shipping

0:43:09thank you very much to ero for

0:43:10sponsoring our show ah you know

0:43:16obviously Swift is open-source I don't

0:43:17know if you're aware that Chris but uh

0:43:18huh i'm looking at yeah right at how do

0:43:21you feel like it's been going so far in

0:43:23i'm assuming similar to John you know

0:43:25I'm gonna take away the obvious answer

0:43:27it's going well but it may be too kind

0:43:29of Q conversation what's been really

0:43:32surprising about going open source and

0:43:35and you can get a specific or as broad

0:43:37as you'd like to be but I got to imagine

0:43:38that that you guys had a bunch of

0:43:41expectations about what going open

0:43:43source would be like and i would imagine

0:43:44some of them are accurate and some of

0:43:45them weren't so how do you feel like the

0:43:47Swift open source project has been going

0:43:48yeah so I think that 22 to summarize

0:43:52it's not just going well it's going far

0:43:54better than we ever anticipated and that

0:43:56caused problems so for context

0:44:02I and many of the other people on the

0:44:03team had quite a bit of open-source

0:44:05experience so elevated claim for example

0:44:08open-source they they were they are

0:44:12vibrant and amazing communities with

0:44:14hundreds of contributors and dozens of

0:44:18companies are collaborating in public

0:44:20and doing really amazing stuff even some

0:44:22of the most staunch competitors you know

0:44:25AMD AMD and intel and AMD and nvidia or

0:44:29you know the these kinds of companies

0:44:30that really you know don't get along on

0:44:33the business side of things but the

0:44:35engineers were great at working engineer

0:44:37to engineer and solving problems and

0:44:39stay focused on making things better so

0:44:42with Swift we assume that would be very

0:44:45similar to that but what we found is

0:44:48that there was a major difference

0:44:49between launching Swift and launching

0:44:51something like playing for example which

0:44:53is that when Swift losses open source

0:44:56which was december third just over a

0:44:58year ago in RI had I don't know a

0:45:02million users and we're playing

0:45:06started and every yond and didn't really

0:45:09care and grew slowly Swift open-source

0:45:13started and was amazingly fast moving

0:45:16changing tons of people hundreds of

0:45:19patches on the first day it was totally

0:45:20crazy and we didn't really know what to

0:45:22do and so I mean it's I i am like super

0:45:26impressed with the the Swift open source

0:45:28community and all the people involved in

0:45:30the energy and just think it's just it's

0:45:32hard it's hard to believe how great

0:45:34everything is but it it really did

0:45:37create problems because we had to figure

0:45:38out things like how we're gonna stay up

0:45:40on in terms of the patch volume

0:45:42okay we're getting hundreds and hundreds

0:45:43of emails a day to the Swift evolution

0:45:45mailing list what are we going to do

0:45:47about that

0:45:47how do we balance apples goals for

0:45:52pushing Swift forward with the goals of

0:45:53the community which are somewhat more

0:45:55chaotic and last directed and I think

0:45:58that over last year we've learned a lot

0:46:00and you can see that reflect in terms of

0:46:02how the project is being managed and how

0:46:05how things are being done but it's

0:46:07always been super important to be not

0:46:10just open source but I guess if you look

0:46:14at an open source project there's

0:46:15there's three kind of easy levels to to

0:46:18look at it the first basic requirement

0:46:21the ssi to be open sources to make your

0:46:23code available right and so you can

0:46:25produce tarballs periodically and that's

0:46:27technically open source you can allow

0:46:31contributions and if you do that through

0:46:33pull requests or patches or whatever

0:46:35then people can engage with it and make

0:46:38the code better

0:46:39there's open community so not only can

0:46:42do you accept a patch but you actually

0:46:44discuss things and talk about things in

0:46:47public and then there's open design and

0:46:50with Swift we went all the way and I

0:46:51think it was totally the right thing to

0:46:53do because we really want to benefit

0:46:54from all the smart people in the world

0:46:56and it was it and still is amazing but

0:47:00it's also very challenging

0:47:01yeah I can imagine so building upon the

0:47:05open sourcing of Swift it was very

0:47:08surprising to me and I think a lot of

0:47:09the community that Apple chose to upload

0:47:12the entire commit history it wasn't just

0:47:14a here's day one initial commitment you

0:47:16know 11 gazillion files and

0:47:18the gazillion lines of code was that

0:47:21plan from the start that you guys would

0:47:23upload the entire history i would say it

0:47:25was planned from the start that we'd

0:47:26upload the entire history but it was

0:47:28kind of assumed from the start that it

0:47:30would be open source at some point

0:47:32interesting and if you look at apple's

0:47:35history i think that's that's in the

0:47:38tool space I think that's fairly

0:47:40straightforward lvm wasn't started at

0:47:44Apple but almost almost everything Apple

0:47:49does end up going public in the same

0:47:51space

0:47:52clang was started at apple and claimed

0:47:54was contributed by Apple to the album

0:47:57project and that included his full

0:48:00revision history other projects like ldb

0:48:02and lip c++ compiler RT and all these

0:48:06other smaller projects were started at

0:48:09apple and contribute to the open source

0:48:10project so Swift it was kind of assumed

0:48:13that that would happen at some point the

0:48:15the real question was when when is the

0:48:17right time for that to happen and if you

0:48:20remember the days of Swift window Swift

0:48:231022 dunno in particular was a super

0:48:26chaotic super learning as we went and

0:48:29trying to keep up with the community and

0:48:30trying to keep up with the pace and the

0:48:32challenges that people were facing and

0:48:34trying to get key pieces like the

0:48:36air-handling model and protocol

0:48:37extensions and all that kind of stuff

0:48:38that it always been planned but one

0:48:41didn't make it into Swift window and so

0:48:43really focusing on that was the priority

0:48:45and so we just decided that open

0:48:48sourcing Swift at 10 for example would

0:48:50be really actually bad for the Swift

0:48:53community because there are major pieces

0:48:54language that weren't there we would

0:48:57have to spend a lot of time community

0:48:59building and try to explain to people

0:49:00the difference between what was not

0:49:03there but should be there and what is

0:49:05you know there and known broken and we

0:49:08had to do a lot of that one swift was

0:49:10eventually open source than in last year

0:49:13right after Swift to toronto but it was

0:49:15a much better place and the team could

0:49:16actually handle the load of doing the

0:49:19community and we didn't expect the kind

0:49:22response we got but we knew that

0:49:23building a vibrant open source community

0:49:25is a lot of work and it's not something

0:49:26to be under underestimated we want to

0:49:29make sure that Swift to was out the door

0:49:31and it wasn't kind of a resource

0:49:33contention between do we finished with

0:49:36two or do we start open source and that

0:49:39and that's why I was announced to be

0:49:41open source by the end of the year

0:49:42because that meant that we can start

0:49:44working on the open-source

0:49:45infrastructure project bringing up all

0:49:47the other stuff that takes to make

0:49:48something open source after Swift is

0:49:50complete so my recollection of the Swift

0:49:54three in the planning stages was that

0:49:56ABI compatibility was somewhere in the

0:50:00list of I'm not gonna call them goals

0:50:03but somewhere in the list of items that

0:50:05were in contention for a swift 3 and at

0:50:07some point along the development process

0:50:09of Swift 3.avi comparability got

0:50:11postponed and given what you just said

0:50:14about the importance of ABI

0:50:15compatibility for the framework people

0:50:17inside apple or whatever

0:50:18what can you say about that is the did

0:50:22you learn lessons from ABI compatibility

0:50:25being planned for the three or do you

0:50:27think like it was never really hard and

0:50:29fast goal and you totally made the right

0:50:31call to delay it or are you disappointed

0:50:33that it didn't make it

0:50:34I'm i think many of those things I mean

0:50:37we're disappointed that it got delayed

0:50:38but it was totally the right time I

0:50:41think that there's a couple of things to

0:50:43learn from that one of which is software

0:50:44schedule planning is hard but breaking

0:50:49this

0:50:50yeah so that that's definitely still

0:50:52thing there's also that we didn't

0:50:55anticipate the so the original plan

0:50:58you're referring to was announced before

0:51:00switch is open source or super sorry was

0:51:02announced the same time so it was open

0:51:03source and so we had no idea that the

0:51:06community would descend on swift and

0:51:08suddenly have all these new demands and

0:51:11consume so much time and energy from

0:51:14that have took away from programming to

0:51:16do the community building peace so that

0:51:18was part of it

0:51:19another part of it is that even

0:51:21stability is super important but it's

0:51:24not as important as people think it is

0:51:26for application developers it's really

0:51:28important to apple but what we realized

0:51:32in the three timeframe is that the thing

0:51:33after developers would benefit from the

0:51:36most was actually sourced ability so who

0:51:40actually wants their application be

0:51:42broken when they get new version of

0:51:43xcode

0:51:44well really nobody right and so and so

0:51:48kind of halfway through the release we

0:51:49pivoted and source can soar stability

0:51:52became the goal and so I'm really

0:51:54excited that ones with the three don't

0:51:56want or so for comes out that I'm still

0:51:59gonna be able to build with three code

0:52:00and even if there are minor changes that

0:52:03need to be made for one reason or

0:52:04another that you can upgrade and you

0:52:07have great compatibility with your old

0:52:09code and you don't have to start the

0:52:11migrator before you can do anything so

0:52:13it's gonna be a great improvement for

0:52:14people's lives

0:52:16alright so moving on a little bit to

0:52:18that kind of this branch topic here of

0:52:21server-side Swift with on linux things

0:52:23like this is is switched on the server

0:52:27something that you know I i know it's

0:52:30possible and there's like some seemingly

0:52:33some early efforts now to get that going

0:52:35it do you see this as like a major goal

0:52:38first trip to the language and you're

0:52:40not not any kind of official capacity

0:52:42just like you as a person who who likes

0:52:44and user Swift and has as a you know a

0:52:46strong role in the community like is

0:52:50server-side Swift a thing that you want

0:52:53to be like a major thing to compete with

0:52:55you know things like Python rails and

0:52:57stuff on the server or is it intended

0:53:00more to be kind of like a specialized

0:53:02thing that is probably not going to be

0:53:03in widespread use by like web and

0:53:05back-end programmers so I think that

0:53:08everybody involved with the project

0:53:11wanted to be the way to do server

0:53:13development and I think that to

0:53:17understand that you have to understand

0:53:18were just kind of entering phase 2

0:53:21Swift's world domination plan so this

0:53:27will probably sound insane but maybe in

0:53:29a few years it will sound less than

0:53:31saying it but the way to look at it the

0:53:34way i look at it and the way that many

0:53:36other people have to look at it is that

0:53:37we want swift be the next big thing

0:53:41right we wanted to be more popular than

0:53:42Java was or is we want to be more

0:53:45popular than Python we want to be maybe

0:53:46even more popular than see someday right

0:53:48that's that would be a stretch but but

0:53:51that would be great and so if if thats

0:53:54problem solved how do you solve that

0:53:55problem well open source is a huge piece

0:53:58of that right without open-source it's

0:53:59not going to be available in other

0:54:00people's platforms it's not gonna be

0:54:02used in education and teaching as that's

0:54:05just not gonna happen and so open

0:54:06sources a major piece of that if you go

0:54:09back farther before that Swift world

0:54:12domination doesn't happen unless with

0:54:14games relevance right and I think Swift

0:54:18has nailed the relevance side of things

0:54:20with the iOS and a platform development

0:54:23side of the story were you know maybe

0:54:27not Marco but their new most people are

0:54:29pushing the Swift very quickly and

0:54:31seemed very happy using actually i

0:54:33started using it today

0:54:34okay good call i switch today in honor

0:54:37of this interview and I wrote my first

0:54:39Swift today in an entire extension and

0:54:42overcast report was written Swift Wow a

0:54:45fantastic you go up proud of you is it

0:54:47is a good day

0:54:48yeah and more and more will be coming

0:54:50although there's a lot of jealousy there

0:54:53so i probably long way away from being

0:54:55all Swift yeah well that's that's okay

0:54:58sweetie loves objective-c exactly the

0:55:01way i look at it is that the first steps

0:55:03with world domination is getting to

0:55:05relevance and 24 language gets relevance

0:55:08that needs a killer app killer app right

0:55:10needs a reason why people would want to

0:55:12use it and for swift that's clearly iOS

0:55:14and mac and other Apple platform

0:55:16development it seconds be open source

0:55:19third it needs to go beyond right and

0:55:24then the next the next step the frontier

0:55:27where it makes sense to expand out to or

0:55:28if you're no real time strategy game

0:55:31player the the natural expansion is is a

0:55:33a place where the language is where the

0:55:39language actually only needs a few

0:55:41things added to it to be good and so I

0:55:44think switch server development is

0:55:45really that next step because for

0:55:46development is very similar to app

0:55:48development in terms of the kinds of

0:55:49code that you need to write it's very

0:55:51similar terms of many of the frameworks

0:55:53the major missing piece is that it needs

0:55:57to run on linux and other pertinent

0:55:59server development platforms in and

0:56:02there's then the whole library ecosystem

0:56:05that needs to be built out

0:56:06and so I think that we're in right now

0:56:07is we're in a really exciting phase

0:56:09where those platforms are being built

0:56:12those frameworks and different pieces of

0:56:15infrastructure coming together and

0:56:17that's that's really the idea behind

0:56:18this with server working group that's an

0:56:21official part of Swift org is getting

0:56:22all the people working in the space to

0:56:24talk to each other and i'm very excited

0:56:26about that i'm going beyond that though

0:56:29you know it with the goal of being more

0:56:32relevant than Java for example if we're

0:56:34going to pick on java is the dream of

0:56:38Swift is to scale the way from scripting

0:56:41language all the way down to low level

0:56:43systems programming like you could write

0:56:45a colonel in it or you write some

0:56:47firmware on it and i think the

0:56:49descripton side of it is maybe more

0:56:51straightforward when when this the

0:56:54community and the core team have time to

0:56:57deal with it because they're talking

0:56:58about in integrating regular expressions

0:57:02integrating multi-line string literals

0:57:04and integrating other other features

0:57:05like that that people love from

0:57:07scripting languages and those are big

0:57:10projects but those are kind of well

0:57:13known i guess and they will fit and drop

0:57:15right into the existing existing system

0:57:19the Swift provides systems programming

0:57:22is I think the big the big next frontier

0:57:25and I think that's where Swift can

0:57:27really distinguish itself from java or

0:57:29other other competitive languages

0:57:32because among other widely used

0:57:35languages out there there's really

0:57:36nothing that can unseat see your C++

0:57:38there's rust for example is an

0:57:42interesting language but hasn't got a

0:57:43lot of adoption yet and I think the

0:57:46Swift is more interesting than rust and

0:57:48some of the high-level application

0:57:50demands but we'll see rust is a great

0:57:53language as well and so I'm really

0:57:56interested to see the swift for systems

0:57:59programming work get going and I think

0:58:02that would be a couple years out before

0:58:04it really comes together but once it

0:58:05does i think the Swift can be really be

0:58:09the next big thing and I'm excited about

0:58:11that

0:58:11so just a second back a second to the

0:58:14server role of it you know possibly

0:58:16taking over there like you know that i

0:58:18would love like you know Here I am using

0:58:19PHP and just see these these two

0:58:22languages that are fairly old and I mean

0:58:24PHP is way more hated than Jesse he but

0:58:27but you know this is the two languages i

0:58:29use for various historical and practical

0:58:32reasons it on the server a lot of what

0:58:35what makes a language succeed or fail on

0:58:39servers and web development is not

0:58:41necessarily how good of a language it is

0:58:43but how how healthy and universal and

0:58:48develop the framework surrounded r44

0:58:50things like web development how

0:58:52practical and easy it is to deploy and

0:58:54run on servers so you know Swiss being a

0:58:58being its own good language that alone I

0:59:02don't think is enough to get it to get

0:59:04traction on the server do like do you

0:59:06think is it necessary for the Swift team

0:59:10like the the core group who's running

0:59:13Swift whether that's Apple or the

0:59:15community of contributors with apple or

0:59:17whatever else is important for them to

0:59:20develop and and kind of make official an

0:59:23official set of server and web

0:59:25frameworks and libraries and tools or do

0:59:29you think it's sufficient to take the

0:59:31risk to let the community do that but at

0:59:33the risk of fragmentation and maybe

0:59:35nothing ever taking off also less

0:59:38interesting question and keep in mind

0:59:40that i don't have a lot of practical

0:59:42experience building server apps so this

0:59:44is not missing much

0:59:45my my my my opinion based on talking to

0:59:49other smart people so the approach this

0:59:52with projects taking is looking to

0:59:54standardize what's colloquially but not

0:59:57correctly known as a server foundation

0:59:57correctly known as a server foundation

1:00:00kind of technologies so it will not be

1:00:02named for foundation but that kind of an

1:00:05idea and so take the encryption the

1:00:07network transport the HTTP kinds of

1:00:09things that that everybody needs for

1:00:12server technologies and standardized

1:00:14that as part of this project but at

1:00:16least in the short term not look to

1:00:18standardize a specific web framework

1:00:20because they're half dozen different

1:00:21completely different approaches that are

1:00:24interesting different different kinds of

1:00:25demands and I don't think there's a

1:00:28clear winner yet it's possible that

1:00:30after the first few rounds shake out

1:00:32there will be an equivalent you know

1:00:35ruby-on-rails kind of dominant pattern

1:00:37and a framework that emerges and if so

1:00:41then maybe that would make sense but I

1:00:43don't think it don't think Swift needs

1:00:45for there to be a winner the high level

1:00:47as far as what can the Swift core

1:00:51language and language and compiler folks

1:00:54do about this

1:00:55I think there's really two things one of

1:00:57which is this what package manager I

1:00:59think the package manager is really key

1:01:01to scaling out the ecosystem and getting

1:01:04people to work together with the common

1:01:08ecosystem it's also the best and i think

1:01:11only way to build swift code on linux

1:01:12and out and other platforms they don't

1:01:15mean to single out linux it's just it's

1:01:17the biggest server development one so i

1:01:21think that's a big piece i think the

1:01:22other piece that there will be really

1:01:24amazing is when the concurrency model

1:01:26comes up because i think that's one of

1:01:27the reasons that makes go particularly

1:01:30compelling for server development is

1:01:32that has really well developed story

1:01:34there along with all the libraries to go

1:01:36together for server and cloud

1:01:37development and so when the pieces the

1:01:40concurrency model's start coming up

1:01:41which will probably the first pieces

1:01:43will ship in Swift five is my guess that

1:01:46will make swift and even more compelling

1:01:48can answer for that kind of space but i

1:01:52don't think i don't think we need a

1:01:54server language feature and that's what

1:01:56you're asking

1:01:57so going to the other extreme from the

1:02:00server-side scripting to gain back the

1:02:03systems programming one thing I've seen

1:02:05related to this and and i think this all

1:02:08stems from

1:02:09with this quote from the original apple

1:02:11Swift book that Swift was designed to

1:02:13scale from hello world to an entire

1:02:15operating system like that that it's the

1:02:17world domination plan was there from the

1:02:18beginning right yep you know absolutely

1:02:20in the first paragraph of the thing when

1:02:23I talk to people now about swift as a

1:02:27systems language often wanna hear is

1:02:30people relating the ways that is not

1:02:34ready to be a systems language yet which

1:02:36is you know what you just talked about

1:02:37but then deciding based on that that

1:02:40because it's not ready to be a systems

1:02:42language now it never will and then they

1:02:44dismiss it like other words like that

1:02:46may have been the plan but clearly were

1:02:50reasons x y&z I you still have to you

1:02:52see for this therefore Swift is just a

1:02:54way to ride my apps for iOS and mac OS

1:02:57and how I mean I'm not sure what you can

1:03:01do with it but do you think about how

1:03:04how to change perception as the language

1:03:07its older and its old age now it's up to

1:03:09three-point out like this

1:03:10it starts at you know that that I nobody

1:03:12puts whipped in the corner like that

1:03:13that people think they know people think

1:03:16they know what Swift is and they say was

1:03:17this language it does this thing and

1:03:19that your plans for world domination may

1:03:21be thwarted by the idea that people are

1:03:23pigeonholing language because it's been

1:03:24around for a few years and I think they

1:03:25know it took that you're onto something

1:03:27I think that's real problem i don't

1:03:29think systems programming is the place

1:03:30where that's a problem yet though

1:03:31because if I don't think it's it's

1:03:34reasonable for me as a crazy man going

1:03:37around saying swift swift will someday

1:03:39do systems programming really well you

1:03:41should you should remember that I don't

1:03:43think that's that's a good way to spend

1:03:44time at the moment when it does that

1:03:47then the question becomes how do you

1:03:49make that make people aware of that and

1:03:53how do you get get people to change

1:03:55their prior misconception but at that

1:03:57point I'm you can a very practical

1:03:59look you can do this and this is so

1:04:00great and now you get all the advantages

1:04:02to see but it's safe then you could

1:04:04point to specific reasons why it's

1:04:06better so I'm not too worried about it

1:04:10in the in the space of systems

1:04:11programming but i think there is a

1:04:12practical from today which is Swift is

1:04:16open-source the Apple development

1:04:19community knows what very well but

1:04:22much everybody else doesn't right and if

1:04:25you go to a random save the next

1:04:30developer just to continue talking about

1:04:32linux they all perceived Swift as being

1:04:35an apple thing and only relevant to

1:04:37apple platforms and they don't even

1:04:39learn about it right and so I think that

1:04:42there is an interesting challenge there

1:04:45and probably the best way to do that is

1:04:47for deep existing communities to keep

1:04:51just growing and as this switch on

1:04:54server community grows and more and more

1:04:56people are successful building and

1:04:58deploying server after and Swift i hope

1:05:01that naturally more people will become

1:05:02aware of that effort will will then find

1:05:06out about Swift maybe try it out find

1:05:09out that it's actually really great and

1:05:10if it's not really great let us know so

1:05:12that people can fix whatever the

1:05:14problems are and then just kind of grow

1:05:16out but i'm i'm actually perfectly okay

1:05:19with the growing a little bit slower

1:05:20than it has been so John asked about you

1:05:24know the span between basically

1:05:26low-level systems programming and

1:05:27high-level applications programming i'm

1:05:29also curious that it seems like 15 /

1:05:32Apple have have presented Swift as not

1:05:35only spanning that wide range which

1:05:37seems impossible but also spanning a

1:05:40wide range of skill levels of programmer

1:05:42so you know there's there's the Swift

1:05:44playgrounds up an ipad where you have

1:05:45total novices writing Swift as their

1:05:47first language all the way up to people

1:05:50writing like you know systems-level

1:05:51things and and complicated apps you were

1:05:54very experienced programmers it do you

1:05:56think is even is it possible for

1:05:59language to do all of these things well

1:06:01and if so do you think Swift has

1:06:04achieved that so I do think it's

1:06:07possible i think it has achieved it for

1:06:10the things that is doing so far and i

1:06:12think the challenges that as new

1:06:14capabilities get added that it doesn't

1:06:16jeopardize that

1:06:17but knowing knowing the the folks

1:06:19involved and the source compatibility

1:06:21guarantees and everything else like that

1:06:22i'm pretty sure that won't be a problem

1:06:24going forward

1:06:25here's the way I look at swift you'll

1:06:28never you're unlikely to run to anybody

1:06:31that says Swift is a simple language

1:06:32right

1:06:35so it is a complicated language it's not

1:06:36simple in the sense that scheme is

1:06:38simple for example it's also not simple

1:06:41in terms of way talk to Jesse developer

1:06:44that the law claim that objective c is

1:06:46simple ignoring all the sea parts i

1:06:48would never claim that for the record

1:06:50alright so see itself as it is a very

1:06:53complicated weird language is just that

1:06:55people don't typically use the weird

1:06:57part right and so that perceives he's

1:07:00being very simple the so the way the way

1:07:05the way i project the Sun Swift is that

1:07:07the the secret to Swift and being easy

1:07:10to learn easy to use as a teaching

1:07:13vehicle but also powerful enough to

1:07:16solve the problems that need to be

1:07:17solved is that the complexities inherent

1:07:19in the language needs to be

1:07:21progressively disclosed so imagine that

1:07:24someday Swift had inline assembly

1:07:25support for example only really low

1:07:28level systems people are really bit

1:07:30twiddling assembly programmers would use

1:07:32this feature everybody else would ignore

1:07:34it and it would not increase the

1:07:35apparent complexity to language of the

1:07:37language to people who didn't know it

1:07:39was for feature like that the most

1:07:40important thing is to make the syntax

1:07:42clear so that when you run up into it

1:07:45you discovered about this code you know

1:07:47that you don't know what it does right

1:07:50and so you say oh that's interesting i

1:07:52don't know what that is and then you can

1:07:54do a search and find out find

1:07:56stackoverflow oracle whatever it is they

1:07:59explain what this is to me and then you

1:08:00can say okay I get it and either care I

1:08:02don't but at that point I'm you can then

1:08:04dive in and and learn and learn more

1:08:07information

1:08:07assembly is never something that a swift

1:08:10playgrounds person working through 12

1:08:12code 3 would ever want to know about

1:08:13right and and i think that is really the

1:08:16this one of the secrets to Swift's

1:08:19success in terms of getting people on

1:08:21board and and being really approachable

1:08:23for people at all levels is that just

1:08:26like in Python you can start with just

1:08:28print hello world you could do the exact

1:08:30same thing and Swift and Swift you don't

1:08:32need the / end on the string it's really

1:08:35just print hello world and from there

1:08:38you can introduce the concepts you can

1:08:40grow it out over time and this really

1:08:41allows people to

1:08:43understand either the syntax language or

1:08:46the conceptual programming things that

1:08:49they're learning as i go and the design

1:08:52of Swift is really geared around that

1:08:54and so what that means if you project on

1:08:56the systems programming features

1:08:57comparing Swift to rest for example I

1:09:00think it's very likely that swill get

1:09:02features for memory ownership control

1:09:05which will allow really really high

1:09:08performance allow solving performance

1:09:10problems with arc for example but unlike

1:09:13rust we can't make that be a core part

1:09:15of the type system that everybody has to

1:09:17use it has to be something that

1:09:18sufficiently smart programmers when

1:09:21they're solving a specific performance

1:09:22problem and abusing or an embedded

1:09:26Colonel programmer might want to use but

1:09:28an application developer can completely

1:09:30ignore and so that's really the

1:09:32challenges its deploying similar type

1:09:35system mechanics and other low-level

1:09:37language geekery to the problem but do

1:09:40so in a way that is tasteful and allows

1:09:42most normal people to just completely

1:09:44ignore everything you just said

1:09:46kind of reminds me of saying I once

1:09:49heard that you're free to adopt

1:09:49respective you would like which is

1:09:51making things easy and make hard things

1:09:53possible we sponsor tonight by audible

1:09:56with an unmatched election of audiobooks

1:09:58original audio shows news comedy and

1:10:01more get a free 30-day trial at

1:10:03audible.com / ATP if you want to listen

1:10:07to it audible has it you can listen to

1:10:08audiobooks from virtually every genre

1:10:10anytime anywhere you can play audible

1:10:12audiobooks on phones tablets computers

1:10:14even Kindles and ipods if you still have

1:10:17them

1:10:18audio books are great for flights long

1:10:20road trips or even your daily commute

1:10:22you might think you don't have time to

1:10:23read books but you'd be surprised how

1:10:25many audio books you can hear each year

1:10:27even only listen to and from work every

1:10:29day that time really adds up

1:10:31audiobooks bring books to life many of

1:10:33them are read by author's themselves

1:10:34which adds an extra dimension to the

1:10:36text and with audible you can take risks

1:10:39and try new authors without regret

1:10:41because audible offers their great

1:10:43listen guarantee if you start an

1:10:45audiobook and don't like it you can

1:10:46exchange it for another one for free so

1:10:49check it out for yourself you can see

1:10:51and listen when you begin your free

1:10:5330-day trial you get your first audio

1:10:55book for free and there's

1:10:56no stress and obligations you can cancel

1:10:58your auto membership at anytime with

1:11:00audiobooks and spoken word audio product

1:11:02you will find what you're looking for

1:11:04get a free 30-day trial by signing up at

1:11:07audible.com / ATP that's audible.com /

1:11:11ATP thanks to audible for sponsoring our

1:11:14show so Chris you brought up just a

1:11:20second ago how simple pythons hello

1:11:22world is and one of the things I've

1:11:25wondered about Swift is it seems to some

1:11:28degree like it's kind of what we call it

1:11:31the Rorschach test of languages in that

1:11:33I came to Swift a sort of its kind of

1:11:36the objective-c but but my day job with

1:11:39it was a c-sharp guy and I see to my

1:11:41eyes anyway and i hope you don't find it

1:11:43insulting but I feel a lot of C sharp i

1:11:45see a lot of C sharp and Swift and I've

1:11:47heard javascript people again I hope

1:11:48that's insulting say oh I see some

1:11:50javascript in here and I presume that

1:11:52John could say that that he sees pearl

1:11:54and I presume Marco could say that he

1:11:55sees PHP in it but hopefully that's not

1:11:57an icy go and I don't see any PHP didn't

1:12:00know you get what I'm driving at though

1:12:02and so so i guess what i'm asking chris

1:12:05is obviously there's a huge objective-c

1:12:07influence and Swift but was there anyone

1:12:10or maybe just a couple of other

1:12:11languages that you feel really really

1:12:13inspired swiftor do you really think

1:12:15it's that it's the melting pot language

1:12:17that one would always hope to create my

1:12:20finger really is a melting pot because

1:12:21we weren't so the goal Swift isn't 28

1:12:25some other language is to just be great

1:12:28right right and so you can take I think

1:12:31that Swift looks or I and the other

1:12:34people in the core team the other people

1:12:35working working on swift have looked at

1:12:38many many other languages and if you're

1:12:40asking nerd there's tons of Pascal

1:12:42concepts built-in protocols really are

1:12:45just like very similar Haskell cons

1:12:47constructor for example it's really

1:12:51about taking the best ideas from where

1:12:53we can get and assembling them together

1:12:55and one of the major goals Swift is for

1:12:57to be familiar feeling right because

1:13:00it's familiar with familiar feeling that

1:13:02makes it easier for people to pick up

1:13:04and they're not fighting

1:13:06unnecessary barriers to just get in the

1:13:09way of adoption and so it being familiar

1:13:11is actually success in that way and this

1:13:15is not it is not a result of Swift

1:13:18trying to be go or C sharp or whatever

1:13:21it's about taking the best ideas from

1:13:23them and assembling them together and I

1:13:25think that you could probably pick a

1:13:27language and there's some some good idea

1:13:29that came from them including you know d

1:13:32or dark or go or whatever I mean Swift

1:13:36does really draw from many many

1:13:38different sources and it's really hard

1:13:40to to really small out so the question

1:13:45that Casey wrote for me to ask you right

1:13:47now LOL how do you know that was me

1:13:50that's definitely you is why should I

1:13:53give a shit about Swift right now and i

1:13:55think im gonna modify the selected that

1:13:58was written it was written for you my

1:14:00friend but yes that is how I you right i

1:14:02figured it's not quite on style

1:14:04no butBut I I you know as as somebody

1:14:07who is just dipping a toe and Swift you

1:14:10know as of 12 hours ago you know it has

1:14:13waited this long you know just to see

1:14:15faithful / hold out / lazy people who

1:14:18don't learn new things very quickly

1:14:20what's kinda like the sales pitch for

1:14:21for adopting swift now I'm I guess two

1:14:26different answers 1 which is you don't

1:14:28have to stop Swift right and I don't

1:14:30think objective-c is going to go away

1:14:31anytime soon

1:14:32Apple sports c and c++ and it doesn't

1:14:35have any specific data there's no

1:14:38obvious benefit of dropping objective-c

1:14:39as well as obviously they have a ton of

1:14:42objective-c code themselves so I i don't

1:14:44think that there's any reason you have

1:14:47to adopt swift swift just tries to be a

1:14:50better answer so that you'll want to

1:14:52adopt it and in terms of Y Swift in its

1:14:57name kind of his connotes a whole bunch

1:14:59of different things including

1:15:00performance but really to me the most

1:15:03important aspect of it is programmer

1:15:04productivity right the the idea that

1:15:06Swift is really optimizing for is not

1:15:08letting you pound out the code as fast

1:15:10as possible

1:15:11it really is designed and optimized for

1:15:14as a programmer you can spend the least

1:15:16amount of time to get to a working

1:15:18program

1:15:19as fast as possible right and so getting

1:15:22to working program includes the time it

1:15:23takes to debug something includes the

1:15:25time it takes to iterate and develop and

1:15:28fight the type system and all the other

1:15:29things that you have to do but that time

1:15:33you spend having to fight the type

1:15:35system or figure out optionals or things

1:15:37like that pays itself back because your

1:15:39program works more often and so you're

1:15:41actually not spending time chasing down

1:15:43dangling pointer bugs or you know and in

1:15:47the case of objective-c you get an

1:15:49unrecognized selector air you know

1:15:51whatever whatever those kinds of bugs

1:15:53are and swift rise to help you by the

1:15:56language guiding you and helping design

1:15:58things so it also has a lot of things

1:16:00that just make common things easier

1:16:02so you can switch on a string for

1:16:04example and then match the different

1:16:05cases you can use enums which are a

1:16:08super genomes or just from the language

1:16:12geekery world is just algebraic data

1:16:13types there are tons of different

1:16:15functional languages

1:16:16they're not novel by any means but

1:16:18they're so useful and they solve so many

1:16:21problems out of the box and just having

1:16:24your fingertips is great so there's

1:16:25there's a lot of things about Swift and

1:16:27probably depends on who you are what

1:16:29your experiences are one of the problems

1:16:31I've seen with people moving from

1:16:32objective-c Swift initially is that they

1:16:34tend to write swiftcode just like they

1:16:36wrote objective-c and while that's a

1:16:38really good way to make a migraine

1:16:41application / or to get familiar Swift

1:16:43it really means that you're missing a

1:16:45lot of the value and Swift and and I

1:16:48think that that's okay it's just it

1:16:50means that you know an objective secret

1:16:52and we're going to swift first starts

1:16:54out by saying I don't see what the big

1:16:55deal is

1:16:56and only after they spend some time and

1:16:58do they start to realize that they are

1:17:00more productive they do feel better they

1:17:02reach for things that they otherwise

1:17:04wouldn't have and I think it's really

1:17:07interesting to talk to people that have

1:17:09done for example your Swift programming

1:17:10but then have to go back to objective-c

1:17:12and to me that's all been there that

1:17:14that's the really interesting and so it

1:17:16was your experience it felt absolutely

1:17:19archaic in a granted on and I like to

1:17:22think of myself as a nice guy and so I'd

1:17:23probably say something like that even if

1:17:24it wasn't true but the reality situation

1:17:26is it felt just so old it just felt old

1:17:31and I don't know

1:17:32i wish i could quantify that better but

1:17:35it just felt like I was back in time in

1:17:38a way that didn't feel good in something

1:17:40like in some ways it was kind of nice so

1:17:42I i do like optionals I understand why

1:17:46their thing I i I'm all in on it but at

1:17:50the same time there are times when I

1:17:51have to do a guard ladder whatever the

1:17:52case may be that a real right that I

1:17:53really would just like to move on with

1:17:55my life and so not having to worry just

1:17:58being being able to be a little more

1:18:00cavalier was kind of nice but the

1:18:03reality of the situation is as usual as

1:18:04you said earlier is that that's a

1:18:06potential hazard down the road and and

1:18:08that's why i like the way that that you

1:18:10and the team of have set up the type

1:18:12system and Swift but gosh going back to

1:18:14objective-c is so frustrating for me

1:18:17because it just feels like I'm going

1:18:18back in time like I'm going backwards

1:18:21yeah and i think that the challenge

1:18:23that's what faces because this focus is

1:18:25on building you know programs that work

1:18:28and that can be maintained and all the

1:18:30other things that we care about for

1:18:32large-scale software it's maybe not the

1:18:35best solution for I want to pound out a

1:18:38quick script use it once and then throw

1:18:41it away right because in that case you

1:18:43don't care about maintenance you you can

1:18:45fit all the code in your head you don't

1:18:46care about good occupation of the other

1:18:49things that Swift encourages and that's

1:18:52that's the case we're having a dynamic

1:18:54type system can be really really

1:18:55liberating and so you know maybe that

1:18:59will be a challenge for some people but

1:19:01i think that as with games ecosystem

1:19:03against libraries its other advantages

1:19:06will be so great that hopefully it will

1:19:09be a good solution for those kinds of

1:19:11tools as well good those kinds of

1:19:12scripts as well so I'm a minute ago you

1:19:15mentioned how gypsy programmers people

1:19:17like me who are new to Swift tend to

1:19:20write Swift like objective-c and and

1:19:23that's that's concerned I've had it it's

1:19:24one of the reasons that i have

1:19:26procrastinated adopting Swift up till

1:19:28now because I have a have a pretty good

1:19:32grasp of objective-c now and I i right

1:19:36objective-c code that i know is

1:19:38basically idiomatic and and soundly

1:19:42designed like I write code

1:19:44in the style of what I think apples code

1:19:46would look like based on the little bit

1:19:47of you know headers that we see I feel

1:19:50like I right good to see code that is

1:19:51kind of correct and an idiomatic leave

1:19:53sound with Swift you know III assumed

1:19:56that the the idioms would not yet be

1:19:59developed for a while and an established

1:20:01for a while and so I didn't want to

1:20:03write a bunch of bad swiftcode it or or

1:20:06like incorrect or or weird swiftcode you

1:20:10know investing heavily in it so early on

1:20:13do you think that it's at a point now

1:20:14where the idioms are starting to get

1:20:17worked out enough that's less of a

1:20:19concern now

1:20:20yeah I do I'm so I think it's definitely

1:20:23fair to say that in the script one this

1:20:25with two timeframes Swift as language

1:20:28was changing really rapidly and I could

1:20:31see why you'd feel like you're on uneven

1:20:32footing and not really sure what the

1:20:34language is much less what the idioms

1:20:35are Swift three really is quite well

1:20:39baked out and so I expect going forward

1:20:41the the new things are going to be

1:20:42additive they're not going to be

1:20:43changing the existing patterns and so I

1:20:46think this with community in general is

1:20:48gone through the same same kind of thing

1:20:50that you're feeling there where it's not

1:20:52really clear the one right way to solve

1:20:55the problem and maybe there are

1:20:56different good ways to do it with

1:20:58different trade-offs won't clear to

1:21:00people but I think that's I think people

1:21:02are starting to understand that no there

1:21:04are a few style books that are being

1:21:06written by people that I've heard about

1:21:07and so when those come out those will

1:21:09hopefully be good and those will be a

1:21:11good resource but another thing to keep

1:21:15in mind is the Swift is opinionated I

1:21:18guess that it is the way to say so it

1:21:21really does encourage you to do the

1:21:22right thing where where it can and so

1:21:24I'm for example if you use bar for

1:21:27everything

1:21:28the this with compiler will say hey you

1:21:31mark this is a bar but it could be a let

1:21:34and that's just it subtle way of and and

1:21:37let me fix it for you right and that's

1:21:38just it subtle way of encouraging you to

1:21:40use and mutable values which is a very

1:21:43small Clinton but it's just kind of

1:21:45pushing you in the way that it it thanks

1:21:48leads to better code and immutability

1:21:50for local variable doesn't matter that

1:21:52much accepted to communicate something

1:21:54more to the person us to read maintain

1:21:56your code

1:21:57and and so I think this really does kind

1:22:00of encourage you down the right lines in

1:22:02some ways but on the other hand and

1:22:04other places where you're saying hey

1:22:06should only be a class should be a

1:22:07struct the train out some more nuanced

1:22:09and it's it's a hard thing in the Swift

1:22:11compiler can't just know what your

1:22:12problem is solved and so it can help you

1:22:15with that you mentioned Swift style

1:22:18guides before as as a creator of a

1:22:22language how how much influence do you

1:22:26do you feel like you want to have over

1:22:28things like style things that really are

1:22:31you know it's not it's not features it's

1:22:34not technically it's just like you know

1:22:36what you will know different style

1:22:38things like oh you know I usually have

1:22:40one exit at the bottom or you should

1:22:41exit early or I mean like there are

1:22:43things you can do even things like

1:22:46spacing embraced out for crying out loud

1:22:48or you know just the sort of idioms of

1:22:50like this little three-line sequences

1:22:52how you do this transformation on this

1:22:53data structure you mentioned you know

1:22:56language features can influence that's

1:22:58obviously as language designer and

1:22:59participate in that process you can

1:23:00effectively guide the community

1:23:04style-wise in one direction or another

1:23:06but eventually you come to the stuff

1:23:08that's just plain opinion and style and

1:23:11several other languages have had fairly

1:23:14concrete styles set forth early on by

1:23:17the creators we're going to go like an

1:23:19RC style type things

1:23:21how much do you care about being able to

1:23:24influence the i guess the personal

1:23:26preference or the aesthetic portions of

1:23:30style that are outside the things that

1:23:32you can influence as part of language

1:23:33design so as with any programmer the has

1:23:38a pulse

1:23:39I have my preferred style and so clearly

1:23:43that's the right way

1:23:44naturally naturally well so i guess more

1:23:48seriously the land yourself does

1:23:49influence some of the things you

1:23:51mentioned so for example it's my

1:23:53personal opinion that this style where

1:23:56you have a single entry single exit from

1:23:58every function i think that leads to

1:23:59really difficult to maintain and read

1:24:01code and that's one of the things that

1:24:03led to the guard statement existing the

1:24:05guard really does encourage you to use

1:24:07early exit so you handle the exceptional

1:24:09cases and they don't have to

1:24:10worried about them so that's that's an

1:24:12example of where the language design is

1:24:14kind of encouraging you to you know have

1:24:17multiple returns from a function other

1:24:20more subjective things like the number

1:24:21of spaces i personally I'm a fan of to

1:24:24space in dense i know some people are

1:24:25staunchly for other people are staunchly

1:24:28ate some people think 80 columns is

1:24:29right answer right i mean all these

1:24:31things are there are actually really

1:24:33good arguments on all sides

1:24:35some people argue that you should use

1:24:37three space and then because then you

1:24:38certainly can't have tabs in your file

1:24:40wow so this is where this it comes down

1:24:45to you have to pick your battles print

1:24:47and people will disagree for a long time

1:24:50and you can make arguments on both sides

1:24:52but the inn where we've come down is

1:24:56that it's just more practical to let

1:24:58people write code they like to look at

1:25:00and that comes at a cost of if you

1:25:02switch between teams or you pick up

1:25:04somebody else's package then maybe it's

1:25:05indented the wrong way but I don't think

1:25:08that really actually actively hurts the

1:25:11comprehensive comprehensibility of the

1:25:13code that's merely annoying and having

1:25:17something like go format which

1:25:19rigorously enforces the one true way I

1:25:21think would have been nice

1:25:23it just wasn't a priority early on and

1:25:25it also just wasn't a battle that we

1:25:26were willing to tackle and I don't know

1:25:29we'll see that I think this is still

1:25:30also an open question in terms of what

1:25:33other tool and gets built and what other

1:25:36things come into the fray because it

1:25:38would be great to have better tooling

1:25:40around this it just hasn't been built

1:25:41out yet and it's a great opportunity for

1:25:43open source as well i was thinking about

1:25:45it the sort of doomsday scenario where

1:25:47like one of these the Swift style guide

1:25:49books comes out and like it makes some

1:25:51like blanket recommendation that like

1:25:53you should you know always use protocols

1:25:56for this and you should never use

1:25:57glasses for that or whatever and it

1:25:58becomes wildly popular accepted as like

1:26:02the Bible of how you rights with

1:26:04programming and you're just some are

1:26:05shaking your head and go no you got it

1:26:07all wrong but I guess you know we'll

1:26:09cross that bridge when we come to it i

1:26:11think at this point you did the language

1:26:13design itself and the communities have

1:26:14enough influence that's not likely to

1:26:15happen but I mean I guess what i was

1:26:17getting at is it do you feel like the

1:26:18you want to write a switch style got

1:26:20like if you had unlimited

1:26:21I'm no I don't think that would be a

1:26:23good use of my time but if I don't look

1:26:25at a time maybe i don't know i started

1:26:27stardancer that theoretical question but

1:26:29but I also get to your concern about

1:26:32somebody came out with a book that was

1:26:33so wrong that would cause me to like you

1:26:36know be really unhappy actually have a

1:26:38ton of faith in the Swift community

1:26:39because if something came out that was

1:26:41so weird i don't think that the

1:26:43community overall would would accept it

1:26:46the community is again just kind of

1:26:49blown away with how beautiful they are

1:26:51and how how many great people have

1:26:54gotten involved with Swift and Swift

1:26:56evolution side of things and it's been

1:26:58really amazing sees people you know they

1:27:00they go through their own personal

1:27:02trajectory on swift mailing list where

1:27:06you know initially they're saying hi I

1:27:08had this idea what you think and the

1:27:11responses know that's a terrible idea

1:27:12for this is a reason but then they come

1:27:16back a little okay how about this you

1:27:18think about this and it was like oh well

1:27:19that's interesting that almost worked

1:27:20but it didn't work because of that and

1:27:22then they come back and say hey well

1:27:24what about this is what we actually

1:27:25tried that and we had that for a while

1:27:27and then had to back out and change this

1:27:28other way because we didn't even think

1:27:29about this other thing and then the next

1:27:31day they're the ones saying this is the

1:27:33way it needs to go and erases you're

1:27:36right we should do that and it's it's

1:27:39it's really hard for people that have

1:27:41been involved in the overall design of

1:27:44anything is complicated Swift to be able

1:27:47to discern the difference between

1:27:48something that is the way it is out of

1:27:51an intentional decision process or the

1:27:54way it is as an actor an accident of

1:27:56history and so as now that the design

1:27:59processes public i think that it becomes

1:28:01a lot easier for people to to tell that

1:28:03and it's very easy for people to kind of

1:28:07understand the difference

1:28:10the one big debate that i haven't

1:28:12actually seen is why is the funky word

1:28:14the funky word and that's one that

1:28:16luckily were now beyond this with three

1:28:18source compatibility threshold so we

1:28:21just can't have that discussion anymore

1:28:23i think i'll be darned

1:28:25but one of the best things of this world

1:28:27open source thing or what

1:28:29the whole evolution process has that

1:28:31like the list of questions like the

1:28:33Aston answers questions like this is a

1:28:35commonly asked question we've dealt with

1:28:36it before and you're probably gonna want

1:28:38to ask it to here's here's about how it

1:28:39was asked before and here's what the

1:28:41answer is not that everybody's gonna go

1:28:43and see that but i love that exists

1:28:45well then it's also hilarious that

1:28:46started out as the commonly proposed

1:28:48list and then at some point it became

1:28:50the commonly rejected list

1:28:52yeah right because it because if it's

1:28:54commonly proposed obviously it's not

1:28:55getting through and so you have to go to

1:28:57be rejected list that you're thinking

1:28:58about asking to change funk to FN you

1:29:01know well my I know you talked about

1:29:03this before but it is well worth

1:29:05touching it again because many people

1:29:07have asked me to ask you this

1:29:11swift and a rough style memory model you

1:29:14you phrased it before as an add-on

1:29:17feature for people who know that they

1:29:20need that type of thing but that it

1:29:22would not like the Swift wouldn't become

1:29:24a language in which all memories manage

1:29:25that way instead this would be a

1:29:26special-purpose feature for particular

1:29:29roles what what roles do you think it's

1:29:33suited for and why why is it not the

1:29:36great memory model for all Swift well so

1:29:39here's here's the way I look at swift in

1:29:42terms of memory management today so

1:29:43Swift is using our it's not using GC if

1:29:46you want to talk about the trailer i was

1:29:48happy to do that thats its own well we

1:29:50can fall into arc i think is overall

1:29:54amazing default for swift because it

1:29:56gives you deterministic destruction it

1:29:58gives you a relatively predictable

1:30:00performance and the ark optimizers able

1:30:03to eliminate most of the reference

1:30:06counting operations that happen

1:30:08the problem that I see with with our at

1:30:11as of today or states with three is that

1:30:15it is not a very controllable model and

1:30:19so if for example you have an inner loop

1:30:22and it turns out the ark optimizers

1:30:24doing a bunch of retains releases and

1:30:26you know that they're not necessary

1:30:28because you know the objects are alive

1:30:30there's nothing you can really do to to

1:30:33solve that problem unless you're willing

1:30:35to drop down to something like the

1:30:37unsafe operation Swift gives you for

1:30:39manipulating references objects the

1:30:41and so it's really kind of satisfying me

1:30:44to have a really safe great memory model

1:30:47that for performance reasons you have to

1:30:48drop down to unsafety right this is a

1:30:51huge cliff and is that also a huge

1:30:53complexity cliff that you have to drop

1:30:55into just solve that problem and that's

1:30:58not really great so the promise the

1:31:01dream is to instead say hey well if you

1:31:03are a sufficiently advanced programmer

1:31:05and you know how this stuff works well

1:31:07what the we can give you options instead

1:31:09of dropping down to unsafe constructs

1:31:12you can actually just use more static

1:31:14type information in annotations in your

1:31:17code and so you add a few annotations

1:31:19say compiler this is ok the compiler can

1:31:22validate those type annotations and

1:31:25check to make sure they're correct so

1:31:26still memory see if it's still

1:31:27guaranteed to be okay but you're paying

1:31:30for it in terms of more of this

1:31:31annotation the code and it's more more

1:31:34heavyweight in terms of coding but when

1:31:37you get out of that is the guarantee

1:31:38that no arc operations are happening

1:31:40once you get to that model then it

1:31:42becomes really interesting because you

1:31:43can say hey well this inner loop i'm

1:31:45going i found in my profile that it

1:31:47really does matter for performance and

1:31:49the ark optimizers doing the wrong thing

1:31:51i can add a few local annotations and

1:31:53now the code is fast it's guaranteed to

1:31:55be fast and the clients of my co don't

1:31:58have to know about it right this is an

1:32:00implementation detail and I think this

1:32:01will really bring us to the best of both

1:32:03worlds where you could have memory

1:32:05safety you can have a progressively

1:32:09disclosed complexity complexities system

1:32:11or most people don't have to know about

1:32:12or think about it but you can get to

1:32:14really low level performance guarantees

1:32:16that otherwise are possible and once we

1:32:19do this i have a feeling there will be

1:32:20certain communities of people that will

1:32:22only want to work in this model and

1:32:24these are the communities that that like

1:32:26rust predominantly so if you're writing

1:32:29a kernel for example maybe don't want to

1:32:30have a reference count your object at

1:32:31all maybe you just want to have you know

1:32:36you say I will always use these

1:32:37annotations my code pervasively and if

1:32:40if that happens then there's a couple of

1:32:42options we can investigate and we could

1:32:43say hey well if you mark this class with

1:32:45some attribute than that get that that

1:32:48prevents the compiler from ever doing

1:32:50reference counting and if it would ever

1:32:51do that it just raises in there

1:32:53or we can even theoretically have a

1:32:55language dialect modes that hey produce

1:32:57produce an heir anytime I did and our

1:32:59cooperation

1:33:00I don't like language dialects and so

1:33:02we'd have to find a way to handle that

1:33:04maybe it would be a warning and then

1:33:06they would upgrade to an error or

1:33:08something like that but I can imagine

1:33:09doing something like that and once we do

1:33:12that then I think you get to a really

1:33:13interesting mode where those systems

1:33:17programming type people that otherwise

1:33:19would be using seriously + + can i use

1:33:21swift and that means they only have to

1:33:24learn one language that means that they

1:33:26could use the same toolset that you know

1:33:28the ID and everything else that works

1:33:30great with Swift it means that they can

1:33:33scale and use the same language for

1:33:36different layers of their stack because

1:33:38many applications have very low level

1:33:41pieces and very high-level pieces and

1:33:42having them written in different

1:33:44languages is really kind of gross

1:33:46I mean I think we've seen this in the

1:33:48objective-c community where one of the

1:33:50best and most annoying things about

1:33:52objective-c is that it has seen it right

1:33:54and so this has been hugely important

1:33:56for objective-c and practice because if

1:33:58you're into a performance problem with

1:34:00obviously message send you can always

1:34:01rewrite that algorithm in see that's

1:34:04really really really important for

1:34:06objective-c being successful and you

1:34:09know both in the days of next on 16

1:34:11acres processors but also today for the

1:34:14low-level cult code the people are

1:34:16writing but also means that that's one

1:34:19of the problems that makes it so that

1:34:20objective c and c are so tightly

1:34:22intertwined and so you can actually take

1:34:24the sea parte out of objective-c without

1:34:26producing a different language who are

1:34:28waiting for swift in the web browser

1:34:30because it's your only Escape speaking

1:34:31of applications that have to have one

1:34:32language at the higher level that i was

1:34:34in the lower one that's about then

1:34:36well there-there are our that that i

1:34:38think is a frontier that is probably the

1:34:40biggest payoff because replacing

1:34:42javascript would be such an amazing

1:34:44service to the world music and art

1:34:49before it's been tried

1:34:51yeah but it's been tried and that's

1:34:52where you know putting my hope on the

1:34:55azan jas the web assembly those kinds of

1:34:57efforts were its enabling languages like

1:35:01Swift and swift swift compiler lv I'm

1:35:04web assembly and Adam jazz both have lvm

1:35:07backends in principle so make it

1:35:08wearable Swift to one of those and go to

1:35:10town with the if those end up winning

1:35:15and they become big big things then

1:35:19maybe that'll be good enough but i don't

1:35:21really know I don't have my crystal ball

1:35:23doesn't go out to the point where

1:35:24javascript is not in web browsers

1:35:26yeah that's the difficult part is not

1:35:28the language mechanics it's the getting

1:35:30a getting something we can understand

1:35:32that in front of as many people as

1:35:33possible because it's also very

1:35:35interested me so again i'm not a web

1:35:37developer and I'm not a I'm definitely

1:35:39not an expert in space but to me it

1:35:41seems the the trajectory of JavaScript

1:35:44has basically made it into an

1:35:46intermediate language it's not something

1:35:48that developers are writing as much

1:35:50directly as they used to

1:35:52and so I c-type script and water other

1:35:53languages that now compiled down to

1:35:55JavaScript and you know they're really

1:35:58using javascript like it like a bit code

1:36:00representation not not a source language

1:36:02that people are are thinking about as

1:36:04much anymore and so I wonder if in five

1:36:06years you know the natural conclusion of

1:36:09that is the javascript is just an adams

1:36:11is type of thing or maybe web assembly

1:36:13does take over but but again I'm not the

1:36:16expert don't know where that directional

1:36:18go i know that some people of JavaScript

1:36:20so we'll see on the topic of using

1:36:25language of the JavaScript to write

1:36:27JavaScript I have to say having done

1:36:29that several times that the tooling why

1:36:32the tooling is not quite where you'd

1:36:34want to be in terms of being able to

1:36:36write in debug and profile in the

1:36:40preferred language and not in the

1:36:42compiled down to JavaScript language but

1:36:45yeah that maybe that'll be like stage 17

1:36:47and this with world domination plan

1:36:49people and I think that's always been

1:36:51true for preprocessors so I mean even

1:36:53see if you do crazy things the macro

1:36:55preprocessor and see you get debugging

1:36:59those things are still or oh yeah and

1:37:02that's one of the reasons that we've

1:37:04been careful about not adding a macro

1:37:05systems Swift early it just brings lots

1:37:08of interesting trade-offs in terms of

1:37:10the debugging and developing and

1:37:11designing experience of 11 last night

1:37:15ABI

1:37:16compatibility according to the timelines

1:37:19that I've heard sketched out that ABI

1:37:22stability could be a thing before async

1:37:27you know our concurrency stuff lands and

1:37:30Swift there any is there any

1:37:31dependencies or chicken egg things

1:37:34involving there like in order to other

1:37:36words in order to do you have to nail

1:37:39down some aspects of your concurrency

1:37:40model before you can commit to ABI

1:37:42stability are they totally orthogonal

1:37:44and I think they're totally orthogonal

1:37:46I'm i think the memory ownership model

1:37:48has more of an impact on ABI stability

1:37:51because there's some very low-level

1:37:52things like how the the getters and

1:37:56setters are code generated for a

1:37:58property for example the end up getting

1:38:01impacted by the ownership model that's

1:38:04one of the reasons that were that this

1:38:0615 is actually actively designing and

1:38:08write it working on a proposal for the

1:38:11memory ownership stuff right now it's

1:38:13not so that it can be a swift for

1:38:15feature it's so that the API stability

1:38:18work can be done in light of the

1:38:20ultimate model for the memory ownership

1:38:22support and I think concurrency is not

1:38:27going to be something that impacts avi

1:38:29stability at all but we'll see in a so

1:38:32than the sufficiently smart programmers

1:38:34that work on swift we'll figure it out

1:38:36and I'm sure it'll be great and I think

1:38:39it's also interesting question of

1:38:40whether ABI stability will be done for

1:38:43swift for because it's a huge amount of

1:38:47work the work as well in a way of course

1:38:49but it's also not clear to me that's

1:38:52really the the right or the most

1:38:55important thing for the Swift community

1:38:57and one of the things that ted has been

1:38:59a huge champion for and I think he's

1:39:02right about is that the most important

1:39:04thing right now is with community is

1:39:05making the compiler more reliable making

1:39:08the error messages better making compile

1:39:11times faster and making a it scale

1:39:14better to large projects and you know

1:39:17we'll see how the rest of the swift for

1:39:18schedule goes but i wouldn't be

1:39:20surprised if at some point they decided

1:39:22that focusing on that is more important

1:39:24than nailing ABI stability and so for

1:39:27so you just mention Ted that's that's

1:39:29Ted kramnik who is that is now the the

1:39:31lead of of the Swift project is that

1:39:33right yeah he's the project lead so ted

1:39:36has been managing the Swift and Clank

1:39:39projects before that for a number of

1:39:41years he's been an apple for 10 years

1:39:43now he's a Stanford PhD graduate he

1:39:47built single almost single-handedly the

1:39:50the first client sack analyzer yeah

1:39:53instead so as they have I love it i'm

1:39:56saying this now as a swift developer you

1:40:00know she should I be like you know as we

1:40:02saw when you announce that you're

1:40:03leaving Apple that we saw some of the

1:40:05people the community asking like she

1:40:06would be worried about his future is

1:40:08Swift in good hands and a minute you

1:40:10know it seems like most of us didn't

1:40:11know Ted at least didn't know his name

1:40:13or or unfamiliar with contributions it

1:40:16does seem like he's been doing quite a

1:40:17bit there anyway

1:40:19should we be worried i don't think you

1:40:21should be worried at all i think swift

1:40:22is an amazing hands as I mean all the

1:40:27things that I am leaving behind an apple

1:40:29and quotes are in really good hands and

1:40:31that's one of the only reason I was

1:40:33comfortable and moving on in the case of

1:40:36Ted Ted is a rockstar he has been

1:40:40running the Swift team and managing them

1:40:42for mean for five years now the roles

1:40:47that we had was that Ted own brand

1:40:50esteem and then report to me and so I

1:40:52would mess around and you know cause

1:40:55trouble by you know trying to get people

1:40:58to do the exciting important thing that

1:41:00I thought and maybe didn't align with

1:41:02the actually important thing and

1:41:04therefore identities chaos into the

1:41:06world but Ted was always the this table

1:41:10runner and the one that was really

1:41:11focused on getting the important things

1:41:13done and he has a number of really

1:41:15exceptional engineers in the team people

1:41:17on the core team like the Gregor John

1:41:19the call Joe Joe graph Dave Abrahams

1:41:22these are all just amazingly brilliant

1:41:25people and the E and also the team

1:41:27includes a bunch of other engineers that

1:41:29are equally dedicated and also just

1:41:32totally amazing people to work with

1:41:34so the Swift team is I think one of

1:41:36apples biggest assets it's it's crazy

1:41:39how brilliant and dedicated there

1:41:41to the project and how excited they are

1:41:43to get to world domination one step at a

1:41:46time coming back to kind of a more broad

1:41:49thing let's say that in this is not

1:41:51meant to be tongue-in-cheek reference to

1:41:53your next job this is just in general

1:41:54you know you're at home you you kids are

1:41:57in bed you know your wife is out or

1:41:59something like that and you want to just

1:42:01write some code you can't use swift and

1:42:03you can't use objective c what are you

1:42:05writing code in and I know that's often

1:42:07dependent on what project you're

1:42:08accomplishing but like what's

1:42:09interesting to you these days that isn't

1:42:11Swift is an objective-c also i mean i

1:42:14have a ton of different interests i mean

1:42:16i really love building great racers for

1:42:18example and have done much stuff with

1:42:20that and I guess I couldn't use Swift

1:42:24with that then I would probably have to

1:42:25use people's pot no also check i was

1:42:29going to call that I wanted to give you

1:42:30a chance to not second-class but what I

1:42:34remind people that were they mentioned

1:42:35you

1:42:36you know what LVL is written in right

1:42:38much applause and not have your your

1:42:41answer bc of loss and and you know how

1:42:43much banks in personal suffering causes

1:42:45to have to be riding c++ all day and

1:42:48building a beautiful world for other

1:42:49people it's just yeah that's where

1:42:52you're entitled down the mines

1:42:54yeah yeah it's it's a pain and suffering

1:42:56that's worth it for the bigger bigger

1:42:58victory that's where the stress comes

1:42:59from years having to use C++ so i could

1:43:01do the same stuff but not feel this much

1:43:03pain

1:43:03that's exactly right i mean there's a

1:43:05lot of suffering that led to switch

1:43:07needing to happen so but but also it

1:43:11really depends i mean maybe if I was

1:43:12doing you know if I was a crazy person

1:43:14ideas pearl for something i don't know

1:43:15from that instead haha well dot but well

1:43:20done terrible to read it though nobody

1:43:23correct

1:43:23hey if it's throw a code then you never

1:43:25need to rewrite haha for John these are

1:43:31as well just wait for my process

1:43:32grammars to land gently and Swift yeah

1:43:35also I mean that's another that there is

1:43:37a great example of pro really has nailed

1:43:40the string processing thing even ants in

1:43:42405 and I think 406 is even better and

1:43:45Swift has a lot to learn from and that's

1:43:47that's something where a

1:43:49so for is really learning from Pearl so

1:43:53that was great

1:43:54oh please don't give him any ammunition

1:43:56this is terrible i was enjoying this

1:43:57interview so much until this very moment

1:43:59look at this way once this happens John

1:44:01will have no excuse

1:44:03oh he will find one nothing is so

1:44:07perfect that it can't be complained

1:44:09about oh my goodness this is a perennial

1:44:13topic on this with Mary lists about

1:44:15dynamic features and Swift you know that

1:44:18our record more dynamic and objective-c

1:44:22and less so in swift swift community

1:44:26there's a lot of angst about when if

1:44:29ever will Swift allow me to essentially

1:44:32you know like what you said before you

1:44:33can write for training in any language

1:44:34you can write a directory name is in any

1:44:36language people will try to write their

1:44:39code in Swift but using models from

1:44:42objective-c that require things like you

1:44:45know it was you know since sending

1:44:48messages objects that you know will

1:44:50receive them but the compiler can

1:44:51guarantee are you know having big

1:44:53heterogenous collections were just like

1:44:54no everything's ID and it'll be fine

1:44:57what I don't know how to frame this but

1:45:00uh I think that the main question has

1:45:03been our dynamic features possible and

1:45:06Swift and are they coming in Swift or is

1:45:09there some kind of philosophical divide

1:45:11where the the most dynamic stuff that

1:45:13people really want will never be

1:45:16forthcoming up for technical reasons but

1:45:17just for philosophical and safety

1:45:19reasons

1:45:20sure so I've been pretty outspoken about

1:45:22this on the switch evolution let's hope

1:45:23this isn't a huge surprise but i don't

1:45:26think there's any tension between

1:45:28dynamic features and Swift the a lot of

1:45:32people want to make a and argue about

1:45:34how it would be impossible to do

1:45:35responders or you can pick your feature

1:45:39the in Swift and the the worst-case

1:45:45scenario frantic blog post I've seen our

1:45:47are things that are along lines of

1:45:50someday Apple rewrite all these

1:45:52frameworks in Swift and then it will be

1:45:54impossible to write first responder

1:45:56chains and then bad things will happen

1:45:58and then we'll be back in the days of

1:46:00MSC or some other bed

1:46:02right i don't even bring that up

1:46:05exactly and so this is this is kind of

1:46:08quite the slope that it's the discussion

1:46:10ends up being I think this blow-up most

1:46:15recently right before WC this year and

1:46:17it was really interesting to wash the

1:46:20the blog post that people are flying

1:46:21around because everybody had a different

1:46:25idea of what dynamic meant happen but

1:46:28none that I'm really so apples not going

1:46:33to counter blog somebody if he didn't

1:46:36like being out somebody's a blog on

1:46:38somebody's post and it's kind of bad

1:46:40form to do that but I thought was very

1:46:41interesting on that none of those people

1:46:43actually came to the Swift evolution

1:46:45list which is the right place to do that

1:46:46and asked about it they all just want to

1:46:49make these inflamed blog posts are very

1:46:51skins so if you get to my opinion my

1:46:54opinion is that none of those are in

1:46:55tension with a swift current design it's

1:46:58just a matter of prioritization and the

1:47:01major problem that Swift has again

1:47:03there's so much work to do there's

1:47:06concurrency model there's actors their

1:47:09systems programming their scripting

1:47:10there's lots of little things that go in

1:47:13there's better better things for static

1:47:16analysis introducing type stay in

1:47:17language there's there there's green

1:47:20postconditions that there's so many

1:47:22things that that will be bricks in the

1:47:25house of Swift that we want to build

1:47:27that need to happen that it doesn't

1:47:30really make sense spent a lot of time

1:47:32building dynamic features so that Apple

1:47:35could theoretically someday do something

1:47:37right

1:47:38were the Swift team is much more focused

1:47:40on the practical necessities of solving

1:47:42the real world problems here now in

1:47:44front of people and dynamic features

1:47:47like many other things will just kind of

1:47:49swap in place when the team s time to

1:47:52design and roll them out and so you you

1:47:55will be able to someday right i expect

1:47:58that you will be able to reflect over

1:48:00all the methods and the class has for

1:48:03example today even the Swift compiler

1:48:05generates all the metadata so you can

1:48:07reflect over all the data members and

1:48:10your and your values we just don't have

1:48:12an API wrapped around that

1:48:13that metadata is what the xcode memory

1:48:16heap viewer uses for example and so

1:48:19there's a lot of that kind of stuff

1:48:20being put in place but just hasn't had

1:48:23time to fully break out yet and when it

1:48:25does come in i think it's gonna be

1:48:26really exciting and will open more doors

1:48:28for new kinds of patterns of people can

1:48:30do and a lot of those will be familiar

1:48:32to objective-c programmers but they're

1:48:34also be super useful for things that

1:48:36people really haven't done an object to

1:48:37see so i don't i don't see a lot of

1:48:40tension there though it's just a

1:48:41prioritization question I think about

1:48:44where your people are coming from is it

1:48:46in terms of prioritization because if

1:48:47you're used to doing things with certain

1:48:49sort of design patterns broadly speaking

1:48:52in objective-c and you want to come and

1:48:53implement the same patterns and Swift

1:48:55and you find that you can't because

1:48:57you're lacking some feature it feels

1:48:59like there's a barrier up in it that

1:49:01students getting back to your writing

1:49:02ready your swift and objective-c style

1:49:04and that's more of a berry they retreat

1:49:06back to breakfast because i can write it

1:49:07the way I want to read it right and it

1:49:08works and objective-c when I try to

1:49:10write it that way and so if it doesn't

1:49:11work for me your choices to change the

1:49:13way you write it to be better fit for

1:49:15swift has or to go back to objective-c

1:49:17I'm so I don't agree with that though

1:49:20because Swift does support all the

1:49:23features for dynamic dispatch and method

1:49:26look up and everything else that

1:49:27objective c does it just requires you to

1:49:29use the objective-c subset of the

1:49:31language because it exposes it through

1:49:33the runtime right that that's that's

1:49:35thatthat's thatthat's the tension

1:49:36because they're like oh I want to write

1:49:38in pure Swift like as if there's

1:49:40something special and magical about not

1:49:42using the objective-c runtime for you

1:49:44know I mean you do get a medal for that

1:49:46oh yeah I guess that makes some people

1:49:49feel good i mean that there are I I

1:49:53can't really count that but I just don't

1:49:55think that it the it's actually fair to

1:49:57say you have to fall back to writing

1:49:59objective-c code no no I got I'm just

1:50:01I'm trying to I didn't write one of

1:50:02these blog posts over to be trying I'm

1:50:03trying I'm trying to express myself that

1:50:05because that was a lot of the concerns

1:50:06like they they want to do it it's they

1:50:08want to do it like the new way and the

1:50:11way and part of doing that is not you

1:50:13know an objective seeing everything like

1:50:14oh that feels like a compromise in the

1:50:15old way so there they are setting their

1:50:18own goal posts for themselves

1:50:19arbitrarily to try to enter into this

1:50:22new world and you know that they they

1:50:25want to jump in with both feet or not at

1:50:27all and and some of them might end up

1:50:28bailing but i think this mean that the

1:50:31way the way look at that is a look at

1:50:34that is an area where the language is

1:50:36not fully baked out it's not fully

1:50:38serving their needs and I think that's

1:50:40really unfortunate right and i'm not

1:50:42going to say that I think that that's

1:50:44not a good thing i want there to be an

1:50:46awesome reflection model I want there to

1:50:47be awesome features for sending messages

1:50:49like there used to just want to be

1:50:52developed and designed in the right way

1:50:54and at the same time they're being

1:50:56annoyed that they can't do those things

1:50:58there are a lot of other people being

1:51:00annoyed about compiled times right

1:51:02ya know I know and and what we have to

1:51:05do is balance between you know just

1:51:08picking two random things where does the

1:51:10engineering effort go and and that's

1:51:13never black and white thing there's

1:51:15never a right answer is all based on gut

1:51:18and judgment and guesses listening to

1:51:21the community talking to people and

1:51:23that's one of the reasons why i think

1:51:24the 15 is really phenomenal is that it

1:51:27really cares that really cares about

1:51:28what people are doing it read you know

1:51:31the various people read the blog post or

1:51:33people on Twitter that like the

1:51:34engagement between us with team and the

1:51:37community is really high because people

1:51:39care and there's no management dictated

1:51:41apple says that you know this engineer

1:51:44should be on Twitter all the time

1:51:45talking to people that's certainly not

1:51:46the case

1:51:47that's because of passion from the

1:51:49developer is because they really care i

1:51:51think that that's the thing that is so

1:51:52amazing about the Swift team is just how

1:51:56important it is in terms of customer

1:51:58service and understanding understanding

1:52:00what people are working through and that

1:52:03insight is what guides the priority

1:52:06decisions and guides the trade-offs that

1:52:08have to be made all the time because

1:52:10nothing really is as clear as you'd like

1:52:13it to be so building on that when all of

1:52:16us were going back and forth and I was

1:52:18tangentially involved with this we're

1:52:20going back and forth auto it switches

1:52:22and dynamic it must be dynamic

1:52:23this is it fair to say that that does

1:52:26eventually make its way inside of apple

1:52:28just and i'm just in that you guys are

1:52:30aware of it that that that's talked

1:52:33about in some way shape or form because

1:52:34it's nice to know and i suspect that it

1:52:37does but it's nice to know that when the

1:52:39community gets perturbed whether or not

1:52:41that that that feeling is justified that

1:52:45at least somebody inside Apple has

1:52:46hurted whether or not they can respond

1:52:48to it at least they've heard it so when

1:52:50we all get our our our pants all twisted

1:52:54up does that does that make it in

1:52:56I absolutely i mean the there are tons

1:53:00of people actually care about what's

1:53:01happening in the developer community and

1:53:03and elsewhere right and that's one of

1:53:05the the things about Apple's approach in

1:53:08terms of not commenting on things that

1:53:09can be frustrating because you don't

1:53:11know but again there's so many people

1:53:14apple that really do care and that

1:53:16sought and you know again I wish those

1:53:20people that were really worried about it

1:53:23would come to Swift evolution and ask

1:53:25because they're exempt discussion about

1:53:26it but that wasn't happening for some

1:53:28reason but certainly there's a ton of

1:53:31people who saw the blogs and i'm sure

1:53:33there are tons of engineers at Apple to

1:53:34feel the same way and wish the

1:53:35priorities were different and I think

1:53:38that the engineering community inside of

1:53:40Apple reflects quite closely to the

1:53:43engineering community outside of Apple

1:53:44as well and i don't think that I hope

1:53:48that's not surprising right angle as

1:53:50many different engineers with different

1:53:52feelings of different opinions and and I

1:53:54think that's a really good thing if if

1:53:56it were a supermodel culture or

1:53:58something like that that probably very

1:53:59unhealthy taking a little bit of a turn

1:54:03so i started writing Swift almost

1:54:06exactly a year ago now and the project

1:54:09that i'm doing at work

1:54:10we're using RX Swift and in

1:54:12quote-unquote reactive programming and

1:54:14i'm curious have you looked into that at

1:54:16all and do you have any particular

1:54:18thoughts on it doesn't seem barbaric to

1:54:20you does it seem clever does it just

1:54:21seem like an odd choice at ya what what

1:54:24is your take on this whole RX thing

1:54:26yeah I i I've definitely seen it i have

1:54:29no real world usage experience for sure

1:54:32most of my opinion is based on reading

1:54:35blogs and reading people talk about how

1:54:37awesome it is well not talking about the

1:54:39problems

1:54:40yeah I from what I gather it seems like

1:54:44it is a way of letting you write a lot

1:54:46less code and code that often works

1:54:49better

1:54:50the first time and i can see that super

1:54:51appealing it also seems like it'd be

1:54:53really challenging to debug it and

1:54:54maintain it and so I don't know between

1:54:56the two proteins between the pros and

1:54:59cons out balances out if I writing a lot

1:55:02of high-level apps i would probably try

1:55:04it out on an opportunity if I liked it

1:55:06but but aside from that are really have

1:55:10a good insight either way if it's the

1:55:12right thing to do or not sure sure

1:55:14absolutely

1:55:15you mentioned just a second ago that you

1:55:16see some problems and you mentioned

1:55:18debugging is there anything else that

1:55:19jumps out at you is being a big concern

1:55:21or or disinterest that it's so different

1:55:23I mean I'm not afraid of different

1:55:26things i think that it's it's more of AI

1:55:30mean the questions I'd raise and this is

1:55:31out of ignorance not out of church or

1:55:33here but you know is what is the

1:55:35community doing is their support line

1:55:37are it is the you know is it changing

1:55:41and evolving in a way that is good or

1:55:43bad

1:55:44and these are just things i don't know

1:55:46because i haven't used to but I have

1:55:48nothing against the RX community at all

1:55:50and I've seen a lot of really cool

1:55:51things I've done and I've read many of

1:55:52the the positive post in RX and and rach

1:55:57native than many of the other things for

1:55:59people using active techniques so and

1:56:01reactive cocoa has also been around for

1:56:03quite a while and at and I think that's

1:56:05a really interesting application that

1:56:07maybe became a lot easier when Swift

1:56:09came out then doing an objective-c

1:56:11absolutely so i think that is great this

1:56:13with can support different kinds of

1:56:15programming models like that because you

1:56:18know if a awesome new right answer to

1:56:20some problem comes out you'd hope that

1:56:22you know that your system can support it

1:56:25and can make it really elegant and

1:56:27beautiful i feel kind of like this is

1:56:30fighting you know the last war instead

1:56:33of the current war but you mentioned

1:56:35garbage collection for his ark bunch of

1:56:37times and obviously that ship has sailed

1:56:39but i would love for you to give a a

1:56:44reasonable summary of what the

1:56:46trade-offs are there because objective-c

1:56:48had

1:56:49garbage collection as you mentioned sort

1:56:51of at yeah it's not there and eventually

1:56:56objective-c dropped the garbage

1:56:57collection and got our kind of course

1:56:59Swift doesn't have garbage collection at

1:57:01all can you talk about the trade-offs

1:57:02there and why Swift is the way it is

1:57:04what would you like a comparison of Arc

1:57:07against the garbage collector

1:57:09objective-c head or garbage collection

1:57:12in theory but you could pick garbage

1:57:14collection in a language perhaps not as

1:57:16burdened as objective-c with see

1:57:18backwards compatibility but you know

1:57:20that's worth mentioning is well i think

1:57:21so i think thats objective-c a decision

1:57:23based on Lovato Lovato had a bunch of

1:57:25implementation concerns and other things

1:57:27that I think don't reflect they don't

1:57:31obviously reflect on to the ultimate

1:57:33garbage collector that you'd ever want

1:57:35to have and so I don't think the

1:57:36comparing against that is interesting

1:57:38but let me turn around so what why do

1:57:40you think our reflections interesting

1:57:41what what is compelling about garbage

1:57:43collection beyond our well it did the

1:57:46idea that

1:57:46yeah memory management is completely out

1:57:48of the hands of the programmer and it's

1:57:50some magical fairy behind-the-scenes

1:57:51will make it all good for you and that

1:57:54you know obviously what you're giving up

1:57:56as you mentioned before that you know

1:57:57even with our cue you lack some amount

1:58:00of control that you'll need to to have a

1:58:01language that becomes a system

1:58:02programming which obviously give that up

1:58:04with the garbage collection because

1:58:05you're not gonna be doing assistant

1:58:06program language in the garbage

1:58:07collected with a garbage collector

1:58:09behind it although i think Microsoft has

1:58:12some project to do that singularity

1:58:13product but they were trying to build a

1:58:14hall West off of garbage collection not

1:58:16quite sure how that's good that's going

1:58:18right but even if even if you ignore the

1:58:20the systems programming side of things

1:58:22because i think it's fair to say a lot

1:58:23of people don't care about that and so

1:58:27garbage collection is proven in the

1:58:28application space for java for example

1:58:32right and pearl but do this reference

1:58:37counting come on

1:58:39oh that's right it's python that users

1:58:40reference counting plus cycle collector

1:58:42so but let me ask you again so you said

1:58:45it means you don't have to think about

1:58:47memory is that true

1:58:49well that's the pitch the pitches that

1:58:51the programmers tend to think about it

1:58:52they have to think about it when the

1:58:53garbage collector and understand the

1:58:55pitch but is it true when the garbage

1:58:57collector starts doing things you don't

1:58:58want and then all of a sudden you do

1:59:00have to think about it but then there's

1:59:01almost nothing you

1:59:02do about it yeah so so here's here's the

1:59:05way I look at it and as you said that

1:59:07the ship has somewhat sale

1:59:08I am totally convinced the artists right

1:59:10way to go up front it is better a whole

1:59:14bunch of different ways it gives you

1:59:16deterministic behavior so you don't get

1:59:19you know it doesn't have the

1:59:21unpredictable stutter problem that

1:59:22people like to bash on gc's the starter

1:59:26problem to me isn't really the issue

1:59:28even though that's what GC haters will

1:59:31bring up all the time it's more about

1:59:33being a reason about when the memory

1:59:35goes away and the most important aspect

1:59:38of that is that are conservative

1:59:40finalized errs so if you use a garbage

1:59:42collector of language use finalized errs

1:59:44finalizer Tsar the thing that gets run

1:59:46when your object is destroyed finalizes

1:59:48have so many problems there are entire

1:59:50that there's entire bodies of work

1:59:53talking about how to work around

1:59:54problems with finalizer for example the

1:59:57finalizer gets around the wrong thread

1:59:58it has get run multiple

1:59:58it has get run multiple

2:00:00x the object get resurrected while the

2:00:02final is running happens non-statistical

2:00:05later you can't count on it and so you

2:00:08can't use it for resource management for

2:00:10database animals and things like that

2:00:12for example there are so many problems

2:00:14with finalizes the artist defines away

2:00:17by having deterministic destruction

2:00:18beyond that art though so the the common

2:00:25that another common thing they're there

2:00:28to two arguments that people make

2:00:29against are in favor of a trace and

2:00:32garbage collector one of which is the

2:00:34arcades overhead because you have

2:00:36retained release operations that run and

2:00:39that is true

2:00:40the other is that you have to think

2:00:44about cycles in arc because it doesn't

2:00:46automatically collects cycles and that

2:00:48is also true and the the rebuttal I'd

2:00:51give to people is that those problems

2:00:54are also true in garbage collection just

2:00:56in different ways so in a garbage

2:00:59collector for example people don't think

2:01:00about it but garbage collection injects

2:01:03additional code into your application

2:01:05just like artists and so you know there

2:01:08are many different garbage collection

2:01:09algorithms not all of them are the same

2:01:10but the most modern garbage collectors

2:01:16that use a nursery for short lifetime

2:01:19objects and then promote them out there

2:01:21generational use something called the

2:01:24right barrier the right barrier is every

2:01:26time you store to a property of an

2:01:30object say you have to run additional

2:01:32code garbage collectors also need the

2:01:34ability to stop all the threads or at

2:01:36least to be able to stop threads at some

2:01:38point in time and they need to be able

2:01:40to do so within a specific time down

2:01:42because they don't want the garbage

2:01:43collector to take forever

2:01:44the artifact that is the typical garbage

2:01:48collectors in Java for example will

2:01:49introduce what's called a safe point

2:01:51into loops and so now in your loops

2:01:55extra code is being run because of the

2:01:57garbage collector and on more aggressive

2:02:00garbage collection algorithms for

2:02:01example i was reading a blog post

2:02:03recently about goes tricolor algorithm

2:02:06they they're touting the advantage of

2:02:09you get really low latency and you get

2:02:11the ability to guarantee response times

2:02:15and more firing level than most garbage

2:02:17collectors but to do that they use this

2:02:19tricolor algorithm which dramatically

2:02:21lowers throughput and that dramatically

2:02:23lower throughput because they're doing

2:02:25almost exactly the same kinds of

2:02:26operations that are kids doing the

2:02:29problem that it then introduces those

2:02:31that these operations that the garbage

2:02:33collectors introducing are sometimes but

2:02:35not nearly as well optimized elazig the

2:02:38ark overhead that the ark optimize it

2:02:40applies to and furthermore there's no

2:02:42kind of out on it so with our I think

2:02:45and hope that the ownership model will

2:02:46give people the ability to take control

2:02:47of those overheads and if it becomes a

2:02:50problem practice or if they're just that

2:02:52kind of person they can take full

2:02:53control over the lifetime of their

2:02:55objects and then know the ark will never

2:02:57happen in the garbage collector you

2:02:58don't have that so the performance side

2:03:01of things I think is still up in the air

2:03:04because are certainly doesn't reduce

2:03:06overhead some of that's unavoidable at

2:03:12least without lots of annotations in

2:03:13your code but I also i think that arc is

2:03:16not done yet a ton of energy has been

2:03:18poured into research for garbage

2:03:20collection particularly since Java has

2:03:23come up there have been hundreds of

2:03:24papers written in the academic circles

2:03:25tons of work and hotspot another other

2:03:29job implementations to do different

2:03:32tweaks and different tunings and

2:03:33different new kinds of albums and

2:03:35garbage collection and that work really

2:03:37hasn't been done for archaea and so I

2:03:39think there's still a big future ahead

2:03:40on the programming side of things the

2:03:43cycle side of things i think it's also

2:03:45really interesting question of how much

2:03:47people think about memory when you when

2:03:51you when I was baiting you a little bit

2:03:53he said anything about garbage

2:03:55collection is that you don't have to

2:03:56think about memory but of course we know

2:03:58that's not true right because if you

2:04:00have a reference to some big object

2:04:02graph they didn't mean to keep around

2:04:03maybe it's in your undo stack then you

2:04:05will quote unquote leak that memory

2:04:07right and that that's true a garbage

2:04:11collector that's true of arc as well

2:04:12it's just any any automatic memory

2:04:15management approach has that problem and

2:04:19so that that is a really interesting

2:04:21thing that there's

2:04:24this question of if you're building a

2:04:26large scale system do you want people to

2:04:29quote-unquote never think about memory

2:04:31do you want them to think about memory

2:04:33all the time like they did in

2:04:35objective-c classic retain and release

2:04:37man retain release or do you want

2:04:39something in the middle and I think the

2:04:41ark strikes a really interesting balance

2:04:42that either whether it's in objective-c

2:04:45orin swift where I look at manual retain

2:04:50really releases being a very imperative

2:04:52style of memory management or Malcolm

2:04:54free all these things where you're

2:04:55telling the code line by line this is

2:04:58where you should do a reference count

2:04:59operation this is where you should

2:05:00release the memory

2:05:02this is what you should do at this point

2:05:03I'm are then takes that model and

2:05:05bubbles it up a big step and it makes it

2:05:07being it makes it be a very declarative

2:05:09model so instead of telling the compiler

2:05:11or the the the telling the compiler that

2:05:15this is the place you should do retain

2:05:17you said say this is an open

2:05:19relationship

2:05:20the cool thing about that to me is that

2:05:22not only does that get rid of the the

2:05:24mechanics of maintaining reference

2:05:26counting and define away tons above by

2:05:28doing that it also means that is now

2:05:30explicit in your code what your

2:05:32intention was and that's something that

2:05:34people who maintain your code benefit

2:05:36from and so by saying that I have a weak

2:05:38point or to the parent object of my my

2:05:41thing that's a really important

2:05:44relationship to know about and as you're

2:05:46looking at the code you maintain the

2:05:47code having that be explicit is very

2:05:49valuable because it talks about the the

2:05:51the relationship between values and to

2:05:55me again with the goal of being ever

2:05:57write large-scale applications and Swift

2:05:59I think that's really useful

2:06:00I also don't think it's hugely

2:06:01burdensome though it is definitely part

2:06:04of the learning curve of learning how

2:06:05swift works that you know has to be

2:06:08balanced in there as well and so i don't

2:06:11know i mean i like it is there are has

2:06:14clear advantages in terms of allowing

2:06:18Swift to scale down two systems that can

2:06:21tolerate having a garbage collector for

2:06:23example if you want to write from where

2:06:24Swift

2:06:25I think it does provide a better

2:06:28programming model where programmers

2:06:29think just a little bit about memory and

2:06:32i think that you know going forward it

2:06:34provides a really high performance model

2:06:37that you can get better than better than

2:06:40garbage collection and almost every way

2:06:41so i think that in terms of trade-offs

2:06:43it's the right one push forward the

2:06:46other I guess third piece that that

2:06:49garbage collection is really bad about

2:06:51is kind of a showstopper for swift which

2:06:54is inner interoperability with C code so

2:06:57if you've ever worked with java or with

2:06:59other similar garbage collector of

2:07:01languages one of the major advantages

2:07:04the garbage collectors give you is they

2:07:05move objects and they need to do that so

2:07:07they can compact those objects so that

2:07:09they can then efficiently doo doo

2:07:11allocations but the problem is that once

2:07:13you start moving objects around if

2:07:15you're interfacing with C code

2:07:17you can't have some random c-code having

2:07:19a pointer to your object and have it

2:07:20moved because then you get a dangling

2:07:22pointer and so once you get down that

2:07:25line you end up with things like J&I the

2:07:27java native interface where you have to

2:07:29explicitly pin things you have to

2:07:31maintain them it's very complicated it's

2:07:34really buggy and are completely defines

2:07:37us away by just saying that ok

2:07:40something's in memory its it has

2:07:42predictable lifetime you can reason

2:07:44about it but provides tools for dealing

2:07:47with unsafe pointers and things like

2:07:48that and that makes the interoperability

2:07:50with existing seco but also with

2:07:52Objective C and maybe someday c++ code

2:07:54really simple really natural and really

2:07:57efficient and I think that's a huge

2:07:58advantage that are provides that really

2:08:01would be impossible to do with with the

2:08:03garbage collector so that's my opinion

2:08:07I mean I think reasonable people just

2:08:09agree obviously but it's something that

2:08:12does come up now and then or used to

2:08:14come up anyway i'm convinced i think you

2:08:15should use our for swift got to get on

2:08:17that

2:08:18yeah Chris thank you so very very very

2:08:21much for coming on the show

2:08:23obviously we are not an interview show

2:08:25is its listeners have discovered over

2:08:27the last hour to but uh but to take your

2:08:30time on what is presumably a very busy

2:08:32time of your life in to spend it with us

2:08:35three knuckleheads I really appreciate

2:08:36it and

2:08:37it was very kind of you to do and

2:08:39hopefully this won't be the last time

2:08:41that we'll hear you want to put on a

2:08:42show or maybe even the show we'll see

2:08:44so thank you Chris I really appreciate

2:08:46it yeah well thank you for having me and

2:08:47I've had fun even though I'm terrified

2:08:49and probably a fool of myself and not at

2:08:51all just me that's that's normal

2:08:53I'm still trying to get over that plus

2:08:55that's our job they're alright thanks

2:08:57lat 43 sponsors this week 0 square space

2:09:00and audible and we will see you next

2:09:02week

2:09:05now the show is over they didn't even

2:09:09mean to be in because it was accidental

2:09:12death was accidental

2:09:16John didn't research Marco and Casey

2:09:20wouldn't let him

2:09:22because it was accidentally was

2:09:25accidental and you can find show know

2:09:29today

2:09:30DP died and it for twitter follow them

2:09:37[Music]

2:09:38yes byl ISS so that's Casey list and a

2:09:43co-pay RM Angie Marco Arment our CC

2:09:50Syracuse

2:09:59what

2:10:04we didn't talk about a woodworking stuff

2:10:09yeah making splitters is fun what about

2:10:12what kind of wood was a table Nina it's

2:10:14it's one that's gonna be tough

2:10:17that's a lot of that's a lot of kinds

2:10:18pretty hard wood right yeah yeah i mean

2:10:21it takes time because it's fun it's one

2:10:23of these one of these areas where you

2:10:25can learn a lot about it over time and

2:10:26keep going deeper and deeper in

2:10:28different spaces and there's never you

2:10:30never know everything so it's a lot of

2:10:32fun to dive into woodworking is also

2:10:34very very different than sitting in

2:10:36front of a computer all day now you're

2:10:38actually going to make something with

2:10:39your hands and you gotta build much

2:10:40higher chance for dismemberment that

2:10:43that is true that is true i think i

2:10:45think it's interesting when programmers

2:10:47have like incredibly analog hobbies

2:10:50it's also letters because the

2:10:52woodworking communities going to this

2:10:53whole cnc thing

2:10:54well they're turning into a very

2:10:57programmed thing again

2:10:58people always say well why don't you get

2:11:00cnc indubitable own like but that's

2:11:02exactly what I'm escaping think I don't

2:11:04want to have to program my little you

2:11:07can't even escape like the community

2:11:09arguments over standards and had to make

2:11:13it so you just cut down a tree in your

2:11:14throat into your workshop and then out

2:11:15comes a table

2:11:16yeah well so that's that's the that's

2:11:18the funny thing about people is that

2:11:19people are all the same they just don't

2:11:21want to notice the greater the effect

2:11:23you know you get two very different

2:11:25communities and they think they're all

2:11:26different but they're actually the same

2:11:27at route so Chris you're obviously

2:11:31changing careers while careers

2:11:34necessarily change employers would you

2:11:36consider yourself a car person I was

2:11:39that's a funny question

2:11:40I mean I obviously need a car cars

2:11:44important but I'm what i consider to be

2:11:47a nun car person I want to accelerate

2:11:49the path to cars being appliances that

2:11:53solve people's problems and i'm

2:11:54personally not the kind of guy who loves

2:11:56doing I'll changes and fiddling around

2:11:59with them I just want something that is

2:12:01reliable that works

2:12:02ideally drives me where I want to go and

2:12:04I don't have to think about it it's just

2:12:07you know solving my problems it's not

2:12:09something I have to care for feed and

2:12:10maintain so but that's kind of the way I

2:12:13look at cars can you drive a stick

2:12:16I did drive a stick before i got my test

2:12:19a few years ago so yeah I it all comes

2:12:22clear now they really from the point of

2:12:23view like it kind of sounds like you you

2:12:25you want to make cars boring and in all

2:12:28in a good weight really and that's a

2:12:30noble goal and I think I can see why you

2:12:33went to Tesla however it is kind of

2:12:35funny that like Tesla is in my opinion

2:12:37clearly the most exciting car company

2:12:40that the world is known for quite some

2:12:41time and to go there

2:12:43well while having a you know such like a

2:12:46an almost nonchalant opinion about cars

2:12:49is interesting to me hehe yeah well I

2:12:51admit that I never considered going to

2:12:54work at a car company before but that's

2:12:57because I always considered car

2:12:59companies to be you know not not in

2:13:03native the kinds of skills that I bring

2:13:05the autopilot i think is really exciting

2:13:07and really big problem and it kind of

2:13:09fits with my my desire solved nearly

2:13:13impulse possible problems and take on

2:13:16new things and so I'm really excited

2:13:18about it so I'm I'm a little curious

2:13:21on that note we've talked on the show

2:13:24before about you-know-what whether we

2:13:26think that that full self-driving cars

2:13:30you know basically when and whether we

2:13:32think those will exist like and what

2:13:33kind of time frame are we talking to

2:13:34decade our lifetimes ever because you

2:13:37know it's one thing to make something

2:13:38that can work on clearly marked highways

2:13:41it's only thing that can make that

2:13:42something work where john lives where

2:13:44there's where there's ice all over the

2:13:45road for most of the year and so like

2:13:47there's no more lane markings there's

2:13:49barely any flat road surfaces or you

2:13:51have all these different ambiguities

2:13:53that you have in in real-world roads do

2:13:57you think you know it again obviously

2:13:59this is not any kind of official

2:14:00capacity as an about to be tested

2:14:02employee do you think big picture that

2:14:07we will see completely self-driving cars

2:14:10capable of driving on effectively all

2:14:11roads within say the next decade and if

2:14:15not maybe within our lifetimes

2:14:17yeah i think that it's very possible

2:14:20within the next decade and hopefully

2:14:22that's a long long long view but keep in

2:14:25mind the goal is to drive better than

2:14:28human

2:14:29and significantly better than him and I

2:14:32think one of the things that gives me

2:14:34confidence in this is that it's not an

2:14:37even playing field because the

2:14:40autonomous cars have more sensors than

2:14:42humans do you know they have they have a

2:14:44radar for example and that means that

2:14:46they're capable of doing things that

2:14:48humans just physically can't

2:14:50and there's a huge technology problem

2:14:53there's huge software problems is not a

2:14:55overnight kind of a thing that you wake

2:14:58up and solve immediately but I I have

2:15:01confidence can be solved certainly

2:15:02within a decade

2:15:03well the good news is it's not gonna

2:15:06take a lot to be better than your

2:15:08average driver from Massachusetts

2:15:09because that's a pretty low bar now

2:15:11because they know how to navigate these

2:15:12intersections like seven different exits

2:15:15all a different angle some of which are

2:15:17one-way none of which are marked so the

2:15:19humans can figure that out it's going to

2:15:21be tough with the sensors family I will

2:15:23drive a Manhattan without any problems

2:15:25I would never drive in Boston oh I amen

2:15:28brother hi completely great

2:15:29alright so bring this back around Chris

2:15:31you made mention why early in the

2:15:32episode that you like yellow cars

2:15:35what ya was awesome

2:15:39my-my-my one of my favorite cars was a

2:15:42yellow mini cooper with black hood

2:15:44stripes on it nice and I make it for

2:15:47convertible it was totally awesome that

2:15:51doesn't get you in the carpool lane but

2:15:52it was a beautiful car in his defense

2:15:56like yellow it like I've seen a like

2:15:58that kind of that kinda category of

2:16:01color like yellow orange maybe teal or

2:16:04maybe even like a pink like the kind of

2:16:06like bright almost neon ish colors can

2:16:09work really well they often don't but

2:16:12then when they work well they look

2:16:14pretty good i could totally see that the

2:16:16only fly see in this plan is that Tesla

2:16:19does not currently offer either yellow

2:16:21as an option or custom paint colors as

2:16:24an option

2:16:24well so now you've broken the real

2:16:26secret they want to hire compiler ok

2:16:28that's why would you gotten your Tesla

2:16:31and yellow if you could have but yes

2:16:33absolutely

2:16:35it was a giant banana

2:16:38yeah and really fully thought the

2:16:43surgeon i don't i don't know if you get

2:16:45that car component you were talking

2:16:47about model s right yeah yeah now i

2:16:49would yellow doesn't seem right for that

2:16:51we're gonna do some shopping mock-ups

2:16:53workshop in a little bit will see that

2:16:55maybe if you got like stripes down the

2:16:56middle

2:16:57yeah I'm if this is the case for details

2:16:59matter so what ya got me that's like it

2:17:02if there's anything if there's any way

2:17:03for you to get like a pet feature

2:17:04implemented in the model line that seems

2:17:07like a pretty good one to do it at least

2:17:09and you know i would ask if I'm sure I

2:17:12know you're probably not gonna be

2:17:12working on on the Bluetooth you I but

2:17:15i'm sure you could probably put in a

2:17:17word with the right people to please

2:17:19improve bluetooth audio but but

2:17:23otherwise if you can't make a yellow one

2:17:25that would be you know in addition to

2:17:27have potentially saving thousands of

2:17:29lives a year with autopilot soft you

2:17:32know getting getting improved bluetooth

2:17:33and yellow option will be pretty

2:17:34impressive

2:17:35well yellow would make are more visible

2:17:36to other drivers are they can drive a

2:17:39bright colors to know that you're there

2:17:40since you don't have any engine sound

2:17:46and i have to say the most effort I've

2:17:49ever done to win over one swift

2:17:51programmer haha so i'm not sure this is

2:17:58scalable approach yeah that's the best

2:18:00strategy forced adoption you should see

2:18:02other me instead of coming on

2:18:03everybody's podcast part of the

2:18:05individual like that's amazing