WEBVTT 00:00:00.000 --> 00:00:00.500 align:middle line:90% 00:00:00.500 --> 00:00:03.210 align:middle line:84% First time it's ever actually worked. 00:00:03.210 --> 00:00:04.470 align:middle line:90% Maybe. 00:00:04.470 --> 00:00:05.496 align:middle line:90% OK. 00:00:05.496 --> 00:00:07.230 align:middle line:90% [LAUGHS] 00:00:07.230 --> 00:00:08.850 align:middle line:84% As mentioned, I am Gary Bernhardt. 00:00:08.850 --> 00:00:11.903 align:middle line:84% On the computer, I look like this. 00:00:11.903 --> 00:00:13.320 align:middle line:84% I own a company called Destroy All 00:00:13.320 --> 00:00:15.660 align:middle line:84% Software that produces screencasts 00:00:15.660 --> 00:00:17.740 align:middle line:90% on various development topics. 00:00:17.740 --> 00:00:20.370 align:middle line:84% And one of the nice things about producing a screencast 00:00:20.370 --> 00:00:24.960 align:middle line:84% is that it takes me maybe 10 hours per week to produce one. 00:00:24.960 --> 00:00:28.470 align:middle line:84% And the rest of the time I spend idly thinking or playing 00:00:28.470 --> 00:00:30.660 align:middle line:84% with side projects just to keep myself interested 00:00:30.660 --> 00:00:33.150 align:middle line:84% and to find new things to talk about. 00:00:33.150 --> 00:00:36.383 align:middle line:84% And this entire talk is the story of one 00:00:36.383 --> 00:00:37.800 align:middle line:84% of those side projects, which I've 00:00:37.800 --> 00:00:41.640 align:middle line:84% been working on for a little over a year, in secret. 00:00:41.640 --> 00:00:45.960 align:middle line:84% And the title of the talk is "A Whole New World." 00:00:45.960 --> 00:00:48.390 align:middle line:84% I'm going to make three confessions 00:00:48.390 --> 00:00:52.590 align:middle line:84% in this talk, the first of which is that I have written 00:00:52.590 --> 00:00:54.960 align:middle line:90% an editor, a text editor. 00:00:54.960 --> 00:00:58.100 align:middle line:90% 00:00:58.100 --> 00:00:59.880 align:middle line:90% Its name is aneditor. 00:00:59.880 --> 00:01:03.390 align:middle line:84% So if I'm at a terminal and I fire up aneditor 00:01:03.390 --> 00:01:06.373 align:middle line:90% on a source file, this is-- 00:01:06.373 --> 00:01:08.040 align:middle line:84% oh, is it act-- yes, it is going, right? 00:01:08.040 --> 00:01:10.620 align:middle line:84% So if I whack Enter, fire up the editor-- 00:01:10.620 --> 00:01:12.370 align:middle line:84% this is editing part of a Twitter client, 00:01:12.370 --> 00:01:15.360 align:middle line:90% another project I wrote for fun. 00:01:15.360 --> 00:01:18.780 align:middle line:84% But it looks a lot like Vim or Emacs or any terminal editor. 00:01:18.780 --> 00:01:19.890 align:middle line:90% It is not like them. 00:01:19.890 --> 00:01:23.397 align:middle line:84% It does a lot of things that they don't do. 00:01:23.397 --> 00:01:24.480 align:middle line:90% Now, it is a modal editor. 00:01:24.480 --> 00:01:27.540 align:middle line:84% I like Vim, as anyone who has seen my screencast knows. 00:01:27.540 --> 00:01:29.700 align:middle line:84% So I can do things like visual line-wise select. 00:01:29.700 --> 00:01:34.170 align:middle line:84% It's keystroke-compatible with Vim in terms of basic editing. 00:01:34.170 --> 00:01:35.850 align:middle line:90% So yes, it's modal. 00:01:35.850 --> 00:01:37.800 align:middle line:84% Yes, it is actually terminal-only, 00:01:37.800 --> 00:01:40.260 align:middle line:84% for reasons that will become clear later. 00:01:40.260 --> 00:01:42.480 align:middle line:90% It is not Vim. 00:01:42.480 --> 00:01:44.520 align:middle line:84% It is much more powerful than Vim. 00:01:44.520 --> 00:01:46.270 align:middle line:90% But it is not an IDE, either. 00:01:46.270 --> 00:01:51.515 align:middle line:84% It's not some giant ball of millions of lines of code, 00:01:51.515 --> 00:01:53.640 align:middle line:84% or however much is in something like Visual Studio. 00:01:53.640 --> 00:01:55.098 align:middle line:84% There's no debugger in this editor. 00:01:55.098 --> 00:01:57.480 align:middle line:84% There are other very powerful features 00:01:57.480 --> 00:01:59.940 align:middle line:84% that are not in other editors, like it 00:01:59.940 --> 00:02:01.973 align:middle line:90% has this idea of layers. 00:02:01.973 --> 00:02:03.390 align:middle line:84% I've been wanting this in my tools 00:02:03.390 --> 00:02:06.240 align:middle line:84% for years, the idea of overlaying 00:02:06.240 --> 00:02:09.400 align:middle line:84% sort of orthogonal information onto the source code. 00:02:09.400 --> 00:02:12.180 align:middle line:84% For example, the sort of classic example of this 00:02:12.180 --> 00:02:16.165 align:middle line:84% is a diff layer or a diff mode in something like Emacs. 00:02:16.165 --> 00:02:18.540 align:middle line:84% If I'm looking at some source code, I whack a single key, 00:02:18.540 --> 00:02:20.490 align:middle line:90% it inlines the diff. 00:02:20.490 --> 00:02:21.420 align:middle line:90% Sort of uninteresting. 00:02:21.420 --> 00:02:23.030 align:middle line:90% Most editors can do this. 00:02:23.030 --> 00:02:24.780 align:middle line:84% But here's a much more interesting layer-- 00:02:24.780 --> 00:02:26.280 align:middle line:90% the crash layer. 00:02:26.280 --> 00:02:28.350 align:middle line:84% So your system is running in production. 00:02:28.350 --> 00:02:31.410 align:middle line:84% It's emitting tons of log files that look like this, maybe. 00:02:31.410 --> 00:02:34.990 align:middle line:84% I'm using Ruby for all my examples. 00:02:34.990 --> 00:02:36.700 align:middle line:84% Sometimes the logs contain crashes, 00:02:36.700 --> 00:02:40.030 align:middle line:84% they contain uncaught exceptions with tracebacks in them. 00:02:40.030 --> 00:02:43.770 align:middle line:84% So if you load these production logs into an editor, it can-- 00:02:43.770 --> 00:02:45.720 align:middle line:84% as you're editing the code, and when 00:02:45.720 --> 00:02:49.260 align:middle line:84% you whack a single keystroke-- it can overlay the trace back 00:02:49.260 --> 00:02:50.200 align:middle line:90% onto the source code. 00:02:50.200 --> 00:02:52.740 align:middle line:84% So every line, all one lines that are yellow here, 00:02:52.740 --> 00:02:55.110 align:middle line:90% were involved in a crash. 00:02:55.110 --> 00:02:57.720 align:middle line:84% And I can narrow onto a single traceback 00:02:57.720 --> 00:02:58.870 align:middle line:90% and move up and down it. 00:02:58.870 --> 00:03:00.330 align:middle line:84% So if I move down this traceback, 00:03:00.330 --> 00:03:03.220 align:middle line:84% the red line is where the crash originated. 00:03:03.220 --> 00:03:05.500 align:middle line:84% So as you're editing the source code, you hit a key, 00:03:05.500 --> 00:03:07.865 align:middle line:84% there's zero delay, and it shows you everything 00:03:07.865 --> 00:03:09.990 align:middle line:84% on the screen that has ever caused or been involved 00:03:09.990 --> 00:03:12.573 align:middle line:84% with a crash, which I think is very useful information that we 00:03:12.573 --> 00:03:13.860 align:middle line:90% don't currently have. 00:03:13.860 --> 00:03:15.330 align:middle line:84% Right now your exceptions are maybe 00:03:15.330 --> 00:03:16.692 align:middle line:90% stored in some web-based system. 00:03:16.692 --> 00:03:18.150 align:middle line:84% You have to open a browser, log in. 00:03:18.150 --> 00:03:21.760 align:middle line:84% It's very cumbersome as a process. 00:03:21.760 --> 00:03:24.800 align:middle line:84% So I've tried to integrate it into the editing experience. 00:03:24.800 --> 00:03:26.340 align:middle line:90% Now, as I go through-- 00:03:26.340 --> 00:03:28.350 align:middle line:84% the first 2/3 or so of this talk is just 00:03:28.350 --> 00:03:30.120 align:middle line:90% explaining this thing to you. 00:03:30.120 --> 00:03:31.870 align:middle line:84% I'm going to show you features, and then I 00:03:31.870 --> 00:03:33.960 align:middle line:84% show you how it works, how the editor does it. 00:03:33.960 --> 00:03:35.400 align:middle line:90% So this one's pretty easy. 00:03:35.400 --> 00:03:37.410 align:middle line:84% For diff layer, it just does a git diff. 00:03:37.410 --> 00:03:38.900 align:middle line:90% Not difficult. 00:03:38.900 --> 00:03:40.620 align:middle line:84% For crash layer, it's a little harder. 00:03:40.620 --> 00:03:42.120 align:middle line:84% It has to parse the logs, of course. 00:03:42.120 --> 00:03:44.880 align:middle line:84% It has to know how to find the tracebacks and read them. 00:03:44.880 --> 00:03:47.880 align:middle line:84% But it also-- the problem is that the production box 00:03:47.880 --> 00:03:49.930 align:middle line:84% might be running a different version of the code. 00:03:49.930 --> 00:03:52.980 align:middle line:84% So if the production box ever printed into the log 00:03:52.980 --> 00:03:55.732 align:middle line:84% what version it was running, what git hash it was running, 00:03:55.732 --> 00:03:57.690 align:middle line:84% then the editor will actually fast-forward that 00:03:57.690 --> 00:03:59.640 align:middle line:84% through time to the current working copy 00:03:59.640 --> 00:04:01.800 align:middle line:84% to figure out what lines are what 00:04:01.800 --> 00:04:03.450 align:middle line:90% so it can display them properly. 00:04:03.450 --> 00:04:06.600 align:middle line:84% Or occasionally it gives up, because it's ambiguous. 00:04:06.600 --> 00:04:09.000 align:middle line:90% It doesn't know. 00:04:09.000 --> 00:04:11.340 align:middle line:84% So layers-- those are some basic layers it has. 00:04:11.340 --> 00:04:13.290 align:middle line:90% It also has more powerful ones. 00:04:13.290 --> 00:04:16.800 align:middle line:84% But I wanted to start with that, because it's a simple feature. 00:04:16.800 --> 00:04:20.220 align:middle line:84% Here is a much deeper and more complex feature. 00:04:20.220 --> 00:04:22.260 align:middle line:84% The editor knows about all the interactions 00:04:22.260 --> 00:04:24.930 align:middle line:84% between all the units of code in your system. 00:04:24.930 --> 00:04:27.020 align:middle line:84% And by this I mean classes in Ruby. 00:04:27.020 --> 00:04:29.250 align:middle line:84% In a language like Clojure, this might be namespaces 00:04:29.250 --> 00:04:32.040 align:middle line:90% and types and things like that. 00:04:32.040 --> 00:04:34.160 align:middle line:84% If I am looking at some source code 00:04:34.160 --> 00:04:36.540 align:middle line:84% and I whack a single keystroke, it 00:04:36.540 --> 00:04:40.920 align:middle line:84% pops up a graph of the interactions 00:04:40.920 --> 00:04:42.510 align:middle line:90% between the classes. 00:04:42.510 --> 00:04:44.922 align:middle line:84% And this is starting at the test for the current file 00:04:44.922 --> 00:04:45.630 align:middle line:90% I was looking at. 00:04:45.630 --> 00:04:49.500 align:middle line:84% Now, I told you this is terminal-only. 00:04:49.500 --> 00:04:51.723 align:middle line:84% So this should immediately make you 00:04:51.723 --> 00:04:54.015 align:middle line:84% ask a lot of questions about, what the hell is going on 00:04:54.015 --> 00:04:54.610 align:middle line:90% on the screen? 00:04:54.610 --> 00:04:57.180 align:middle line:90% 00:04:57.180 --> 00:04:59.970 align:middle line:84% Allow me to cause you to have more questions. 00:04:59.970 --> 00:05:01.800 align:middle line:84% I can actually scroll through this graph. 00:05:01.800 --> 00:05:02.740 align:middle line:90% It's interactive. 00:05:02.740 --> 00:05:06.390 align:middle line:84% So you just saw me scroll from the database spec at the top 00:05:06.390 --> 00:05:07.290 align:middle line:90% down into the tweet. 00:05:07.290 --> 00:05:09.510 align:middle line:84% If I whack Enter on that, I'm navigating the system 00:05:09.510 --> 00:05:10.080 align:middle line:90% dynamically. 00:05:10.080 --> 00:05:12.210 align:middle line:84% Tweet has no outgoing interaction, 00:05:12.210 --> 00:05:15.390 align:middle line:84% so it's a value object at the sort of root of the graph-- 00:05:15.390 --> 00:05:19.270 align:middle line:84% or, excuse me, at the leaves of the graph. 00:05:19.270 --> 00:05:20.770 align:middle line:84% If I switch it to incoming edges, 00:05:20.770 --> 00:05:22.300 align:middle line:84% then you can see all the things that reference tweet. 00:05:22.300 --> 00:05:23.850 align:middle line:84% So I'm just sort of like dynamically 00:05:23.850 --> 00:05:26.430 align:middle line:84% navigating through the system, observing 00:05:26.430 --> 00:05:28.920 align:middle line:90% how the objects interact. 00:05:28.920 --> 00:05:32.460 align:middle line:84% So how does it do that, especially in terminal? 00:05:32.460 --> 00:05:35.880 align:middle line:84% Well, in terms of the data, we need to retrieve and store 00:05:35.880 --> 00:05:36.867 align:middle line:90% the graph information. 00:05:36.867 --> 00:05:38.700 align:middle line:84% You might have seen that and assumed that it 00:05:38.700 --> 00:05:39.783 align:middle line:90% was doing static analysis. 00:05:39.783 --> 00:05:41.850 align:middle line:84% There is no static analysis in this editor. 00:05:41.850 --> 00:05:45.378 align:middle line:84% I think static analysis is not the wrong thing 00:05:45.378 --> 00:05:46.920 align:middle line:84% for dynamic languages, but I've never 00:05:46.920 --> 00:05:48.587 align:middle line:84% seen an implementation that convinced me 00:05:48.587 --> 00:05:50.490 align:middle line:90% that it was really useful. 00:05:50.490 --> 00:05:53.580 align:middle line:84% Instead, it's auditing all the test runs to learn 00:05:53.580 --> 00:05:55.080 align:middle line:90% how the objects interact. 00:05:55.080 --> 00:05:58.470 align:middle line:84% So whenever it runs tests, or actually, 00:05:58.470 --> 00:06:00.480 align:middle line:84% whenever it forks any process, it 00:06:00.480 --> 00:06:02.610 align:middle line:84% has an environment variable called RUBY_AUDIT, 00:06:02.610 --> 00:06:03.630 align:middle line:90% if it's in Ruby mode. 00:06:03.630 --> 00:06:05.970 align:middle line:84% In other languages, the variable changes names. 00:06:05.970 --> 00:06:10.920 align:middle line:84% This points to a FIFO, a named pipe on the disk. 00:06:10.920 --> 00:06:12.360 align:middle line:84% In your tests, you're going to say 00:06:12.360 --> 00:06:15.960 align:middle line:84% RAudit.audit!, which is going to enable the entire auditor. 00:06:15.960 --> 00:06:18.720 align:middle line:84% That's going to fire up the profiler, turn it on. 00:06:18.720 --> 00:06:20.590 align:middle line:84% It's going to turn on a trace function, 00:06:20.590 --> 00:06:22.740 align:middle line:84% so it observes every line as it executes. 00:06:22.740 --> 00:06:25.210 align:middle line:84% It's going to set an at_exit handler for the process 00:06:25.210 --> 00:06:28.290 align:middle line:84% so that when the process exits, all the profile data and all 00:06:28.290 --> 00:06:31.350 align:middle line:84% the tracing data get dumped out to that FIFO that was named 00:06:31.350 --> 00:06:33.505 align:middle line:90% by that environment variable. 00:06:33.505 --> 00:06:35.130 align:middle line:84% And so it's going to dump out a profile 00:06:35.130 --> 00:06:36.680 align:middle line:84% database, a trace database, and it's 00:06:36.680 --> 00:06:39.720 align:middle line:84% going to dump out an unstyled dot file, a Graphviz dot file. 00:06:39.720 --> 00:06:43.890 align:middle line:84% That's the format that I use for the object graph. 00:06:43.890 --> 00:06:45.840 align:middle line:84% So that's just retrieving the data. 00:06:45.840 --> 00:06:48.703 align:middle line:84% To query it, for the simple case like the one I showed you, 00:06:48.703 --> 00:06:50.370 align:middle line:84% all we have to do is take that dot file, 00:06:50.370 --> 00:06:52.410 align:middle line:84% throw some styles in, and push it into Graphviz, 00:06:52.410 --> 00:06:54.550 align:middle line:90% so that's easy. 00:06:54.550 --> 00:06:56.910 align:middle line:84% The rendering itself-- fortunately, Graphviz 00:06:56.910 --> 00:06:59.600 align:middle line:84% takes dot data on standard in, emits images on standard out. 00:06:59.600 --> 00:07:01.755 align:middle line:84% So I just fork off a Graphviz, throw data into it, 00:07:01.755 --> 00:07:04.090 align:middle line:84% read data out, and everything's good. 00:07:04.090 --> 00:07:05.100 align:middle line:90% It needs to be fast. 00:07:05.100 --> 00:07:07.530 align:middle line:84% So when I initially bring that up, 00:07:07.530 --> 00:07:08.960 align:middle line:84% it takes 200 to 400 milliseconds. 00:07:08.960 --> 00:07:11.790 align:middle line:84% Then I fork off a prerendering process that 00:07:11.790 --> 00:07:15.160 align:middle line:84% will prerender all the movements I could make through the graph, 00:07:15.160 --> 00:07:18.100 align:middle line:84% so if I do make those movements, then they're already there-- 00:07:18.100 --> 00:07:21.550 align:middle line:84% sort of like prefetching in a web browser. 00:07:21.550 --> 00:07:23.040 align:middle line:90% So that's how it works. 00:07:23.040 --> 00:07:25.183 align:middle line:84% And there are a lot of useful questions 00:07:25.183 --> 00:07:26.850 align:middle line:84% you can answer with something like this. 00:07:26.850 --> 00:07:29.340 align:middle line:90% Like, I'm writing a web app. 00:07:29.340 --> 00:07:32.880 align:middle line:84% A request comes in and hits this piece of the controller layer. 00:07:32.880 --> 00:07:34.770 align:middle line:90% What code gets executed? 00:07:34.770 --> 00:07:36.660 align:middle line:84% That's a question that is not that 00:07:36.660 --> 00:07:39.060 align:middle line:84% easy to answer with most current tools, 00:07:39.060 --> 00:07:41.850 align:middle line:84% just seeing the graph of what actually happens. 00:07:41.850 --> 00:07:43.410 align:middle line:90% Or what code does this test run? 00:07:43.410 --> 00:07:46.290 align:middle line:84% That's actually even less common to be able to do. 00:07:46.290 --> 00:07:49.710 align:middle line:84% Or maybe the best one, the most interesting one-- what code 00:07:49.710 --> 00:07:51.325 align:middle line:90% might hit this crash point? 00:07:51.325 --> 00:07:52.950 align:middle line:84% So we already have the production logs. 00:07:52.950 --> 00:07:54.650 align:middle line:84% We know where crashes originated. 00:07:54.650 --> 00:07:57.150 align:middle line:84% We can back out a graph of all the code that might have come 00:07:57.150 --> 00:07:59.010 align:middle line:90% into there from other places. 00:07:59.010 --> 00:08:01.080 align:middle line:84% And we can even organize it, maybe, 00:08:01.080 --> 00:08:03.990 align:middle line:84% with the left-hand side being code that did cause a crash, 00:08:03.990 --> 00:08:06.535 align:middle line:84% the right-hand side being code that didn't, all the paths 00:08:06.535 --> 00:08:08.910 align:middle line:84% coming into that one line, and why did some of them cause 00:08:08.910 --> 00:08:09.930 align:middle line:90% a crash and some didn't? 00:08:09.930 --> 00:08:12.210 align:middle line:90% We can ask questions like that. 00:08:12.210 --> 00:08:16.337 align:middle line:84% None of these are implemented, but the data is all there. 00:08:16.337 --> 00:08:18.420 align:middle line:84% But of course, the big question is how in the hell 00:08:18.420 --> 00:08:20.420 align:middle line:84% did I just draw a raster graphics in a terminal? 00:08:20.420 --> 00:08:22.650 align:middle line:90% 00:08:22.650 --> 00:08:24.210 align:middle line:84% Well, you will remember that there 00:08:24.210 --> 00:08:25.740 align:middle line:84% were three confessions I was going to make, 00:08:25.740 --> 00:08:27.032 align:middle line:90% and I've only made one of them. 00:08:27.032 --> 00:08:28.440 align:middle line:90% So I've written an editor. 00:08:28.440 --> 00:08:30.336 align:middle line:90% I've also written an terminal. 00:08:30.336 --> 00:08:34.440 align:middle line:90% [LAUGHTER] 00:08:34.440 --> 00:08:36.230 align:middle line:84% So let's, like, rest for a moment. 00:08:36.230 --> 00:08:39.789 align:middle line:84% And I'm going to just rant about terminals for just a minute. 00:08:39.789 --> 00:08:41.309 align:middle line:84% This is my configuration for iTerm2. 00:08:41.309 --> 00:08:43.226 align:middle line:84% It's the terminal I use in all my screencasts, 00:08:43.226 --> 00:08:46.190 align:middle line:84% and the one I used before I wrote one. 00:08:46.190 --> 00:08:49.350 align:middle line:84% It's currently configured for xterm. 00:08:49.350 --> 00:08:52.070 align:middle line:90% xterm is basically doing VT100. 00:08:52.070 --> 00:08:57.210 align:middle line:84% VT100 is also ECMA-48, ISO 6429, ANSI X3.64, 00:08:57.210 --> 00:08:58.390 align:middle line:90% or just ANSI escape codes. 00:08:58.390 --> 00:09:01.000 align:middle line:84% These all mean the same thing, pretty much. 00:09:01.000 --> 00:09:03.000 align:middle line:84% And it all goes back to this piece of hardware-- 00:09:03.000 --> 00:09:07.452 align:middle line:84% Digital Equipment Corporation, VT100 terminal, 1978. 00:09:07.452 --> 00:09:09.660 align:middle line:84% This is where the terminal protocol we're still using 00:09:09.660 --> 00:09:11.940 align:middle line:90% comes from. 00:09:11.940 --> 00:09:14.370 align:middle line:84% To digress even further, here's a graph 00:09:14.370 --> 00:09:18.057 align:middle line:84% you've all seen before of transistor counts and CPUs. 00:09:18.057 --> 00:09:19.140 align:middle line:90% At the right, you'll see-- 00:09:19.140 --> 00:09:21.140 align:middle line:84% I don't think you can read the words very well-- 00:09:21.140 --> 00:09:23.010 align:middle line:84% but you'll see an Ivy Bridge CPU. 00:09:23.010 --> 00:09:24.540 align:middle line:90% That's a current Intel CPU. 00:09:24.540 --> 00:09:27.120 align:middle line:84% 1.4 billion, with a B as in boy-- 00:09:27.120 --> 00:09:29.810 align:middle line:84% 1.4 billion transistors in that thing. 00:09:29.810 --> 00:09:31.770 align:middle line:84% A P4 is the fourth one from the right. 00:09:31.770 --> 00:09:36.510 align:middle line:84% You can see it's basically a 0 compared to an Ivy Bridge. 00:09:36.510 --> 00:09:41.915 align:middle line:84% So if we scope this just to the P4, the P4 has-- 00:09:41.915 --> 00:09:42.540 align:middle line:90% what is that?-- 00:09:42.540 --> 00:09:44.700 align:middle line:84% 40-something million transistors. 00:09:44.700 --> 00:09:47.460 align:middle line:84% And you can see that a 486 basically has 0. 00:09:47.460 --> 00:09:49.680 align:middle line:84% So let's remove Pentium, scope it to 486. 00:09:49.680 --> 00:09:52.590 align:middle line:84% And you can see at the left there, the 8008 and the 8080 00:09:52.590 --> 00:09:54.030 align:middle line:90% basically have 0 transistors. 00:09:54.030 --> 00:09:58.250 align:middle line:84% They both have on the order of 10,000, I think. 00:09:58.250 --> 00:10:02.800 align:middle line:84% The 8080 CPU, the one with 0 transistors compared to a 486, 00:10:02.800 --> 00:10:03.880 align:middle line:90% is this. 00:10:03.880 --> 00:10:06.723 align:middle line:90% And this is the CPU in a VT100. 00:10:06.723 --> 00:10:08.140 align:middle line:84% You can see it's in a DIP package. 00:10:08.140 --> 00:10:10.432 align:middle line:84% I'm pretty sure I could hand-solder this even if I were 00:10:10.432 --> 00:10:14.020 align:middle line:90% drunk, and it would be fine. 00:10:14.020 --> 00:10:16.210 align:middle line:84% This is the CPU that powers the machine that 00:10:16.210 --> 00:10:18.580 align:middle line:84% defined the terminal protocol that you all 00:10:18.580 --> 00:10:20.860 align:middle line:90% use every single day. 00:10:20.860 --> 00:10:22.135 align:middle line:90% And I find this-- 00:10:22.135 --> 00:10:24.010 align:middle line:84% I don't necessarily think old things are bad. 00:10:24.010 --> 00:10:25.990 align:middle line:90% I love UNIX, and UNIX is old. 00:10:25.990 --> 00:10:29.118 align:middle line:84% But this I find pretty offensive. 00:10:29.118 --> 00:10:31.660 align:middle line:84% And so the things I wanted, the reason I wrote this terminal, 00:10:31.660 --> 00:10:34.100 align:middle line:84% is I wanted raster graphics on the grid. 00:10:34.100 --> 00:10:35.230 align:middle line:90% So it's still a terminal. 00:10:35.230 --> 00:10:36.397 align:middle line:90% It's still a character grid. 00:10:36.397 --> 00:10:38.997 align:middle line:84% But sometimes a character is a blob of raster data. 00:10:38.997 --> 00:10:41.080 align:middle line:84% So it's still easy to think about writing software 00:10:41.080 --> 00:10:41.440 align:middle line:90% for this. 00:10:41.440 --> 00:10:43.540 align:middle line:84% You don't have to think about arbitrary pixel positioning 00:10:43.540 --> 00:10:44.445 align:middle line:90% and views and stuff. 00:10:44.445 --> 00:10:45.820 align:middle line:84% But you can throw raster graphics 00:10:45.820 --> 00:10:49.240 align:middle line:84% in for simple things like the interaction graph. 00:10:49.240 --> 00:10:50.320 align:middle line:90% 24-bit color. 00:10:50.320 --> 00:10:52.090 align:middle line:84% If you've ever tried to get 256 working, 00:10:52.090 --> 00:10:53.890 align:middle line:84% you'd know that stepping into the '90s 00:10:53.890 --> 00:10:56.560 align:middle line:90% is very difficult in 2012. 00:10:56.560 --> 00:10:58.480 align:middle line:90% So I want to try to fix that. 00:10:58.480 --> 00:11:00.380 align:middle line:84% Italics and bold and underline-- basic text 00:11:00.380 --> 00:11:04.150 align:middle line:84% stuff that the original Macintosh had in the '80s. 00:11:04.150 --> 00:11:06.850 align:middle line:84% Momentary keypresses-- like I whack Control and then release 00:11:06.850 --> 00:11:07.350 align:middle line:90% it. 00:11:07.350 --> 00:11:09.142 align:middle line:84% I want to be able to see that in my editor. 00:11:09.142 --> 00:11:10.933 align:middle line:84% I didn't show this feature, but it actually 00:11:10.933 --> 00:11:12.970 align:middle line:84% has a feature where if you hit a keystroke, 00:11:12.970 --> 00:11:16.180 align:middle line:84% it opens a split for the definition of a function 00:11:16.180 --> 00:11:17.170 align:middle line:90% under your cursor. 00:11:17.170 --> 00:11:19.545 align:middle line:84% Then you let go of the key, and the definition goes away. 00:11:19.545 --> 00:11:23.950 align:middle line:84% So it's a very quick way to just see sort of offhand data. 00:11:23.950 --> 00:11:26.020 align:middle line:84% And finally, I don't like mice at all, 00:11:26.020 --> 00:11:29.613 align:middle line:84% but some people do, so it does support the mouse. 00:11:29.613 --> 00:11:31.030 align:middle line:84% So that is why I wrote a terminal. 00:11:31.030 --> 00:11:33.012 align:middle line:84% And now let's go back to the editor 00:11:33.012 --> 00:11:34.720 align:middle line:84% and see some of the more powerful layers. 00:11:34.720 --> 00:11:36.928 align:middle line:84% I showed you the simple diff layer, the slightly more 00:11:36.928 --> 00:11:38.190 align:middle line:90% complex crash layer. 00:11:38.190 --> 00:11:40.867 align:middle line:84% There is a performance layer, which is really cool. 00:11:40.867 --> 00:11:43.450 align:middle line:84% So you're looking at some source code, you whack a single key. 00:11:43.450 --> 00:11:46.605 align:middle line:84% It overlays, on the left, performance information. 00:11:46.605 --> 00:11:48.730 align:middle line:84% Most of those lines have a gray marker on the left. 00:11:48.730 --> 00:11:50.460 align:middle line:84% That means that it was in the bottom 95 00:11:50.460 --> 00:11:52.480 align:middle line:90% percent of the profile. 00:11:52.480 --> 00:11:53.673 align:middle line:90% Yellow means top 5 percent. 00:11:53.673 --> 00:11:55.840 align:middle line:84% Red, which is the very last line of the source code, 00:11:55.840 --> 00:11:57.650 align:middle line:90% means top 1 percent. 00:11:57.650 --> 00:11:59.440 align:middle line:84% So single keystroke, no wait time, you 00:11:59.440 --> 00:12:03.000 align:middle line:90% can see profiling information. 00:12:03.000 --> 00:12:04.750 align:middle line:84% It also has a series of type layers, maybe 00:12:04.750 --> 00:12:07.030 align:middle line:84% the most interesting for a conference like this. 00:12:07.030 --> 00:12:08.200 align:middle line:90% Remember, this is Ruby. 00:12:08.200 --> 00:12:11.710 align:middle line:84% But with a single keystroke, I can overlay type annotations 00:12:11.710 --> 00:12:15.250 align:middle line:90% onto every single function. 00:12:15.250 --> 00:12:17.050 align:middle line:84% It is not doing type inference here. 00:12:17.050 --> 00:12:19.270 align:middle line:84% Remember, there's no static analysis. 00:12:19.270 --> 00:12:22.940 align:middle line:90% This is by observing the system. 00:12:22.940 --> 00:12:25.030 align:middle line:84% And I can also shorten identifiers. 00:12:25.030 --> 00:12:26.262 align:middle line:90% So this is the shorten layer. 00:12:26.262 --> 00:12:27.970 align:middle line:84% At this point, as soon as I turn this on, 00:12:27.970 --> 00:12:30.790 align:middle line:84% the code becomes read-only within a line. 00:12:30.790 --> 00:12:32.920 align:middle line:84% But now it looks quite a lot more 00:12:32.920 --> 00:12:35.260 align:middle line:84% like maybe Haskell would look, or something, where 00:12:35.260 --> 00:12:37.585 align:middle line:84% you can glance at the type information that tells you 00:12:37.585 --> 00:12:38.710 align:middle line:90% what the function is doing. 00:12:38.710 --> 00:12:42.010 align:middle line:84% You can see the structure of it because the names are short. 00:12:42.010 --> 00:12:44.180 align:middle line:90% And finally, a third type layer. 00:12:44.180 --> 00:12:46.148 align:middle line:84% It can annotate every line with the type 00:12:46.148 --> 00:12:47.440 align:middle line:90% of the expression on that line. 00:12:47.440 --> 00:12:50.450 align:middle line:84% Just sometimes sort of useful to see that. 00:12:50.450 --> 00:12:53.170 align:middle line:84% And then when I turn all these off, code goes back to normal. 00:12:53.170 --> 00:12:54.370 align:middle line:90% The code never changed. 00:12:54.370 --> 00:12:58.065 align:middle line:84% Only the representation on the screen changed. 00:12:58.065 --> 00:12:59.440 align:middle line:84% So how does it do all that stuff? 00:12:59.440 --> 00:13:02.325 align:middle line:84% Well, actually, this one's not that hard to explain. 00:13:02.325 --> 00:13:04.450 align:middle line:84% The performance layer just loads all those profiles 00:13:04.450 --> 00:13:07.450 align:middle line:84% that were dumped out by the auditor during test runs. 00:13:07.450 --> 00:13:08.950 align:middle line:84% But of course, you don't want to see 00:13:08.950 --> 00:13:10.350 align:middle line:90% profiling data for unit tests. 00:13:10.350 --> 00:13:11.690 align:middle line:90% That doesn't make any sense. 00:13:11.690 --> 00:13:14.110 align:middle line:84% So if you tell the editor that my perf tests live 00:13:14.110 --> 00:13:16.990 align:middle line:84% in this directory, it will only use those for the perf tests. 00:13:16.990 --> 00:13:19.600 align:middle line:84% And in fact, you can write a single perf test 00:13:19.600 --> 00:13:22.810 align:middle line:84% for the thing you're curious about, scope the layer to that, 00:13:22.810 --> 00:13:25.990 align:middle line:84% turn the layer on, run the test, and with a couple 00:13:25.990 --> 00:13:29.260 align:middle line:84% of seconds of time you can see the performance data 00:13:29.260 --> 00:13:33.510 align:middle line:84% for the test you just wrote inline with your source code. 00:13:33.510 --> 00:13:35.510 align:middle line:84% The type layer is also pretty easy to implement. 00:13:35.510 --> 00:13:38.560 align:middle line:84% You just load up that trace data that the auditor dumped. 00:13:38.560 --> 00:13:42.100 align:middle line:84% And it sounds expensive, but it's not. 00:13:42.100 --> 00:13:44.710 align:middle line:84% Type inference is expensive because there's 00:13:44.710 --> 00:13:46.000 align:middle line:90% a graph it has to crawl. 00:13:46.000 --> 00:13:48.310 align:middle line:84% But if you've just recorded every single call point 00:13:48.310 --> 00:13:50.650 align:middle line:84% in the whole test run, in every test run, 00:13:50.650 --> 00:13:52.690 align:middle line:84% it's a linear scan of that to find out what 00:13:52.690 --> 00:13:55.620 align:middle line:90% the types of the function are. 00:13:55.620 --> 00:13:58.120 align:middle line:84% Finally, you can turn all these on at the same time 00:13:58.120 --> 00:13:59.270 align:middle line:90% if you want. 00:13:59.270 --> 00:14:01.360 align:middle line:84% And it makes it kind of hard to read. 00:14:01.360 --> 00:14:03.238 align:middle line:84% But the most magical part of this 00:14:03.238 --> 00:14:04.780 align:middle line:84% is that if I go into visual line mode 00:14:04.780 --> 00:14:07.197 align:middle line:84% and select those bottom three lines, that bottom function, 00:14:07.197 --> 00:14:10.810 align:middle line:84% and delete it, it marks it as deleted, 00:14:10.810 --> 00:14:12.822 align:middle line:90% but it's still being rendered. 00:14:12.822 --> 00:14:14.530 align:middle line:84% So all of these things can work together, 00:14:14.530 --> 00:14:15.863 align:middle line:90% and you can still edit the code. 00:14:15.863 --> 00:14:18.393 align:middle line:84% You can't edit within a line with shorten layer on. 00:14:18.393 --> 00:14:20.560 align:middle line:84% But with all the other layers, you can actually even 00:14:20.560 --> 00:14:23.410 align:middle line:90% edit within a line. 00:14:23.410 --> 00:14:26.770 align:middle line:84% So assuming-- so this system has all this knowledge 00:14:26.770 --> 00:14:29.470 align:middle line:84% about the structure of your system, the crashes that 00:14:29.470 --> 00:14:30.870 align:middle line:90% have happened in production. 00:14:30.870 --> 00:14:33.886 align:middle line:90% 00:14:33.886 --> 00:14:36.680 align:middle line:84% If I release this editor today publicly, 00:14:36.680 --> 00:14:39.820 align:middle line:84% and it supported your favorite dynamic language, 00:14:39.820 --> 00:14:40.797 align:middle line:90% would you use this? 00:14:40.797 --> 00:14:42.130 align:middle line:90% Do you think you would use this? 00:14:42.130 --> 00:14:44.000 align:middle line:90% Do you think you might use this? 00:14:44.000 --> 00:14:45.590 align:middle line:90% OK, a lot of hands. 00:14:45.590 --> 00:14:48.310 align:middle line:90% I'm very glad to hear that. 00:14:48.310 --> 00:14:53.590 align:middle line:84% I'm also a little disappointed, because-- 00:14:53.590 --> 00:14:55.960 align:middle line:84% I haven't made my third confession yet. 00:14:55.960 --> 00:14:57.440 align:middle line:90% I wrote an editor. 00:14:57.440 --> 00:14:59.170 align:middle line:90% I wrote an terminal. 00:14:59.170 --> 00:15:02.550 align:middle line:90% I also told you an lies. 00:15:02.550 --> 00:15:04.620 align:middle line:84% I did not actually write an editor. 00:15:04.620 --> 00:15:07.810 align:middle line:90% 00:15:07.810 --> 00:15:09.638 align:middle line:90% I did not write a terminal. 00:15:09.638 --> 00:15:11.680 align:middle line:84% Everything you just saw was actually all Keynote. 00:15:11.680 --> 00:15:13.894 align:middle line:84% That was all text boxes in Keynote. 00:15:13.894 --> 00:15:16.798 align:middle line:90% [LAUGHTER] 00:15:16.798 --> 00:15:20.186 align:middle line:90% [APPLAUSE] 00:15:20.186 --> 00:15:22.440 align:middle line:90% 00:15:22.440 --> 00:15:22.940 align:middle line:90% Yeah. 00:15:22.940 --> 00:15:26.050 align:middle line:84% That's a lot of hours of my life you just watched go by. 00:15:26.050 --> 00:15:27.540 align:middle line:84% It takes a long time to fake that. 00:15:27.540 --> 00:15:31.410 align:middle line:90% 00:15:31.410 --> 00:15:35.460 align:middle line:84% Now, the reason I just did this is 00:15:35.460 --> 00:15:38.880 align:middle line:84% that I tried to design a talk demonstrating something 00:15:38.880 --> 00:15:41.310 align:middle line:84% where if I were sitting in the audience who saw it, 00:15:41.310 --> 00:15:43.980 align:middle line:84% I would be, number one, excited that this existed. 00:15:43.980 --> 00:15:45.960 align:middle line:90% This is what I want to exist. 00:15:45.960 --> 00:15:48.000 align:middle line:84% But number two, I would actually be surprised 00:15:48.000 --> 00:15:49.680 align:middle line:90% that someone had built it. 00:15:49.680 --> 00:15:51.830 align:middle line:84% Now, the editor I would find fairly surprising. 00:15:51.830 --> 00:15:53.830 align:middle line:84% But the terminal I would really find surprising, 00:15:53.830 --> 00:15:55.747 align:middle line:84% because the terminal is infrastructure that we 00:15:55.747 --> 00:15:59.190 align:middle line:90% don't really replace, ever. 00:15:59.190 --> 00:16:01.110 align:middle line:84% And now I want to quickly ask the question, 00:16:01.110 --> 00:16:04.230 align:middle line:84% with my small amount of remaining time, why would that 00:16:04.230 --> 00:16:06.690 align:middle line:84% be surprising to me, and why was it maybe surprising 00:16:06.690 --> 00:16:09.480 align:middle line:84% to you, to see that someone had not only written an editor 00:16:09.480 --> 00:16:13.500 align:middle line:84% but rewritten the entire terminal layer? 00:16:13.500 --> 00:16:15.500 align:middle line:84% I think there are two big reasons 00:16:15.500 --> 00:16:16.750 align:middle line:90% that this would be surprising. 00:16:16.750 --> 00:16:19.050 align:middle line:84% And the first is that we have this shipping 00:16:19.050 --> 00:16:23.585 align:middle line:84% culture that is poisonous to infrastructure replacement. 00:16:23.585 --> 00:16:25.710 align:middle line:84% You have people like Seth Godin saying things like, 00:16:25.710 --> 00:16:26.340 align:middle line:90% "Ship often. 00:16:26.340 --> 00:16:28.080 align:middle line:90% Ship lousy stuff, but ship. 00:16:28.080 --> 00:16:29.288 align:middle line:90% Ship constantly." 00:16:29.288 --> 00:16:31.830 align:middle line:84% --which I think is a great way to achieve short-term business 00:16:31.830 --> 00:16:34.110 align:middle line:84% gains, and a great way to discover 00:16:34.110 --> 00:16:35.670 align:middle line:90% what your customer wants. 00:16:35.670 --> 00:16:38.102 align:middle line:84% But when you're a programmer and you are the customer, 00:16:38.102 --> 00:16:39.810 align:middle line:84% and you're writing the system, and you've 00:16:39.810 --> 00:16:41.730 align:middle line:84% been using these tools for 20 years, 00:16:41.730 --> 00:16:43.808 align:middle line:84% you don't need customer discovery. 00:16:43.808 --> 00:16:46.350 align:middle line:84% You need to go off and sit in a hammock for a couple of years 00:16:46.350 --> 00:16:49.650 align:middle line:84% and think hard, which I know has come up at this conference 00:16:49.650 --> 00:16:52.140 align:middle line:90% before. 00:16:52.140 --> 00:16:54.510 align:middle line:84% The second reason I think this is sort of surprising 00:16:54.510 --> 00:16:59.130 align:middle line:84% is that we have not only all this legacy, like the VT100, 00:16:59.130 --> 00:17:01.010 align:middle line:90% or autoconf-- 00:17:01.010 --> 00:17:03.234 align:middle line:84% I don't like saying that word, but-- 00:17:03.234 --> 00:17:04.650 align:middle line:90% [LAUGHS] 00:17:04.650 --> 00:17:07.700 align:middle line:84% We have not only the legacy, but we have a paralysis around it. 00:17:07.700 --> 00:17:10.020 align:middle line:84% We can't even imagine replacing it. 00:17:10.020 --> 00:17:12.082 align:middle line:90% I mean, VT100, 1978. 00:17:12.082 --> 00:17:14.040 align:middle line:84% A lot of the people in this room weren't alive. 00:17:14.040 --> 00:17:16.170 align:middle line:90% I wasn't alive in 1978. 00:17:16.170 --> 00:17:18.750 align:middle line:84% And probably very few of the people in this room 00:17:18.750 --> 00:17:22.150 align:middle line:84% were actively programming in 1978. 00:17:22.150 --> 00:17:25.240 align:middle line:84% So these things have existed forever for us, right? 00:17:25.240 --> 00:17:28.440 align:middle line:84% We can't see the ways in which they're bad, a lot of times, 00:17:28.440 --> 00:17:31.120 align:middle line:90% or we just don't think about it. 00:17:31.120 --> 00:17:33.450 align:middle line:84% So the way that thinking about this editor started 00:17:33.450 --> 00:17:35.730 align:middle line:84% was, I had all these ideas for an editor. 00:17:35.730 --> 00:17:39.510 align:middle line:84% I had them in my personal Wiki, sort of in a Wiki page. 00:17:39.510 --> 00:17:41.230 align:middle line:84% And I realized pretty early on that 00:17:41.230 --> 00:17:43.110 align:middle line:84% you would have to write a terminal to build this editor. 00:17:43.110 --> 00:17:44.818 align:middle line:84% Or it would not be in the terminal, which 00:17:44.818 --> 00:17:47.370 align:middle line:84% is unacceptable to me, because Control-Z 00:17:47.370 --> 00:17:51.202 align:middle line:90% is a thing that I want. 00:17:51.202 --> 00:17:53.160 align:middle line:84% So I realized I would have to write a terminal. 00:17:53.160 --> 00:17:55.243 align:middle line:84% And I thought about, what does that really entail? 00:17:55.243 --> 00:17:59.070 align:middle line:84% Well terminals are coupled to the kernel, via ioctls. 00:17:59.070 --> 00:18:02.370 align:middle line:84% There's a lot of ioctls that the terminals have to use. 00:18:02.370 --> 00:18:05.580 align:middle line:84% Or, well, on both sides of the terminal, ioctls happen. 00:18:05.580 --> 00:18:07.640 align:middle line:84% There's the tty and pty subsystems. 00:18:07.640 --> 00:18:09.940 align:middle line:90% Those are part of the kernel. 00:18:09.940 --> 00:18:13.560 align:middle line:84% So now I'm basically thinking about making kernel changes 00:18:13.560 --> 00:18:19.200 align:middle line:84% to support my terminal, which seems very difficult in today's 00:18:19.200 --> 00:18:19.860 align:middle line:90% world, right? 00:18:19.860 --> 00:18:20.843 align:middle line:90% You can't go up to-- 00:18:20.843 --> 00:18:22.760 align:middle line:84% you can't submit a patch to the Linux kernel-- 00:18:22.760 --> 00:18:25.330 align:middle line:84% well, we're changing the nature terminal, guys. 00:18:25.330 --> 00:18:25.830 align:middle line:90% Here you go. 00:18:25.830 --> 00:18:27.060 align:middle line:90% Here's the patch. 00:18:27.060 --> 00:18:28.545 align:middle line:90% It will not be merged. 00:18:28.545 --> 00:18:30.870 align:middle line:90% [LAUGHTER] 00:18:30.870 --> 00:18:33.810 align:middle line:84% So I'm not exactly sure what the answer to this 00:18:33.810 --> 00:18:39.080 align:middle line:84% is, but I think that it is basically embrace and extend. 00:18:39.080 --> 00:18:40.830 align:middle line:84% I think you build a new terminal protocol, 00:18:40.830 --> 00:18:42.205 align:middle line:84% you make it backwards-compatible, 00:18:42.205 --> 00:18:44.970 align:middle line:84% and you just throw pty's and tty's away. 00:18:44.970 --> 00:18:46.740 align:middle line:84% But no one does things like that, right? 00:18:46.740 --> 00:18:48.870 align:middle line:90% No one papers over-- 00:18:48.870 --> 00:18:53.140 align:middle line:84% really replaces the low-level infrastructure. 00:18:53.140 --> 00:18:56.940 align:middle line:84% The second thing that I've-- this is something I've observed 00:18:56.940 --> 00:18:59.760 align:middle line:84% from other people, is that talking about this actually 00:18:59.760 --> 00:19:03.250 align:middle line:84% makes some people mad, like mad-mad. 00:19:03.250 --> 00:19:05.250 align:middle line:84% They are actually angry that I will 00:19:05.250 --> 00:19:07.525 align:middle line:84% talk about what a new editor might look like, 00:19:07.525 --> 00:19:09.150 align:middle line:84% or what a new terminal might look like, 00:19:09.150 --> 00:19:12.570 align:middle line:84% or what a new UNIX might look like, which doesn't make sense 00:19:12.570 --> 00:19:13.425 align:middle line:90% to me. 00:19:13.425 --> 00:19:15.300 align:middle line:84% And some other people just think it's stupid. 00:19:15.300 --> 00:19:18.360 align:middle line:84% They think it's masturbatory to talk about this kind of stuff. 00:19:18.360 --> 00:19:23.440 align:middle line:84% That part I understand, at least, but I don't agree with. 00:19:23.440 --> 00:19:25.637 align:middle line:84% So if you're in this camp of, why 00:19:25.637 --> 00:19:27.720 align:middle line:84% would we worry about the low-level infrastructure? 00:19:27.720 --> 00:19:31.950 align:middle line:84% Why would we do full redesign of low-level things that have 00:19:31.950 --> 00:19:33.870 align:middle line:90% existed for 40 years?-- 00:19:33.870 --> 00:19:36.240 align:middle line:84% I'd like to present this hypothetical timeline 00:19:36.240 --> 00:19:38.610 align:middle line:84% of a piece of infrastructure being developed, where, 00:19:38.610 --> 00:19:42.330 align:middle line:84% in spring of 2005, someone decides to work full-time 00:19:42.330 --> 00:19:45.240 align:middle line:84% on this thing, to redesign a piece of infrastructure 00:19:45.240 --> 00:19:47.430 align:middle line:90% we've had for a very long time. 00:19:47.430 --> 00:19:50.490 align:middle line:84% For the first year, they just think and prototype, 00:19:50.490 --> 00:19:53.543 align:middle line:90% until spring of 2006. 00:19:53.543 --> 00:19:55.710 align:middle line:84% And then in spring of 2006, they start implementing. 00:19:55.710 --> 00:19:58.380 align:middle line:84% And in fall of 2007, they release this thing 00:19:58.380 --> 00:20:00.810 align:middle line:84% that they've designed for 2 and 1/2 years 00:20:00.810 --> 00:20:02.190 align:middle line:90% without it being public. 00:20:02.190 --> 00:20:04.033 align:middle line:90% They release it into the world. 00:20:04.033 --> 00:20:05.700 align:middle line:84% Does anyone know-- I assume someone must 00:20:05.700 --> 00:20:07.200 align:middle line:90% know what this timeline is. 00:20:07.200 --> 00:20:09.120 align:middle line:90% [INAUDIBLE] 00:20:09.120 --> 00:20:10.380 align:middle line:90% Oh, good guess, Git. 00:20:10.380 --> 00:20:11.630 align:middle line:90% I hadn't even thought of that. 00:20:11.630 --> 00:20:13.153 align:middle line:90% No, this is actually Clojure. 00:20:13.153 --> 00:20:14.070 align:middle line:90% This is the timeline-- 00:20:14.070 --> 00:20:15.840 align:middle line:84% I emailed Rich to make sure I had my dates right. 00:20:15.840 --> 00:20:17.673 align:middle line:84% This is the timeline of Clojure development. 00:20:17.673 --> 00:20:21.310 align:middle line:84% 2 and 1/2 years before it was public, and a full year of that 00:20:21.310 --> 00:20:24.680 align:middle line:84% was just thinking and prototyping. 00:20:24.680 --> 00:20:26.750 align:middle line:84% And I think this is kind of what you 00:20:26.750 --> 00:20:31.010 align:middle line:84% have to do to redesign infrastructure, right? 00:20:31.010 --> 00:20:34.040 align:middle line:84% Rich didn't just like pile some crap on top of Lisp. 00:20:34.040 --> 00:20:38.317 align:middle line:84% He took the entire Lisp and moved the whole design at once. 00:20:38.317 --> 00:20:39.650 align:middle line:90% And that's something you can't-- 00:20:39.650 --> 00:20:42.620 align:middle line:84% at least, I don't think you can get by doing 00:20:42.620 --> 00:20:43.640 align:middle line:90% incremental development. 00:20:43.640 --> 00:20:46.182 align:middle line:84% I think you really have to step back and think hard and spend 00:20:46.182 --> 00:20:50.210 align:middle line:84% a lot of time laying in a hammock, staring at the sky. 00:20:50.210 --> 00:20:53.823 align:middle line:84% So I realize this was a really weird talk. 00:20:53.823 --> 00:20:55.490 align:middle line:84% When Alex invited me to do a lunch talk, 00:20:55.490 --> 00:20:57.290 align:middle line:84% I decided to do something totally crazy. 00:20:57.290 --> 00:21:00.260 align:middle line:84% And usually my talks are not giant lies. 00:21:00.260 --> 00:21:01.070 align:middle line:90% [LAUGHTER] 00:21:01.070 --> 00:21:04.510 align:middle line:90% So I am Gary Bernhardt. 00:21:04.510 --> 00:21:07.040 align:middle line:84% I own Destroy All Software, which produces screencasts. 00:21:07.040 --> 00:21:09.050 align:middle line:84% And I thank you guys very much for letting 00:21:09.050 --> 00:21:10.760 align:middle line:90% me lie to you for 30 minutes. 00:21:10.760 --> 00:21:14.110 align:middle line:90% [APPLAUSE] 00:21:14.110 --> 00:21:20.000 align:middle line:90%