Platform of the future

September 11, 2011

I remember years ago seeing a talk by the YouTube founders that has stuck with me for its simple insight. They said, roughly, that lots of people had webcams or cell phone cameras, lots of people had Flash players for video, and lots of people had the bandwidth to stream video; these three factors would clearly converge into a future of personalized television using the web. YouTube may have been a video piracy website for some while but its mark on history at this point is clearly just what the founders had envisioned.

With that in mind, let me point out some converging factors I see among the programmer circle I hang out in:

The above factors produce an interesting consequence: most smart programmers I know have OS X machines but don't actually care about OS X beyond it being Unixy enough and the hardware working. If you look over the shoulder of Brad Fitzpatrick or Ryan Dahl you see them running a browser and a terminal. I've heard the Plan9 / Go language guys run MacBooks just to run X11.

And the curious thing about this is, at the level these people use the system at, OS X is actually kind of terrible. The hardware is amazing, of course. But the file system is slow; simple things, like making ls show colors, used to require downloading binaries from a third party² installing a compiler requires a 4gb Xcode download; installing now-standard tools like git requires poking around on web pages and unpacking dmgs; all of the core development tools (gcc 4.2, gdb 6) are relatively ancient and slow by a Linux standard. If anything, the attributes that make OS X unique are liabilities to be overcome, not benefits.

One amusing instance of this effect was pithily noted by a coworker: "Chrome is a Windows application developed on Linux and designed on Macs." It's true: when writing cross-platform Chrome code you're best off writing it on Linux where the tools are better, but past that point there's little point in investing many resources into Linux-specific code as there are no users there. (I've read that Firefox is in a similar balance.)

Here are some potential futures:

  1. Developing software on your local hardware becomes obsolete. This appears to be the vision espoused by Google ("the cloud!"). I am skeptical the next Ruby on Rails is gonna be written by someone using EC2 as their development environment but it is maybe plausible.

  2. Apple remains/becomes significant enough that everyone becomes locked into their platforms, regardless of how much they like it. Writing OS X / iOS native software starts sounding appealing to even me once 90% of college grads have MacBooks and open checkbooks. I will be sad if my profession ever gets to the state where you pay 30% of your revenue to the company that owns the platform, but the market gets to decide that.

  3. Apple remains/becomes significant enough that their platforms are the platform of choice for discriminating hackers. Many would say this has already happened — maybe if Brad had started today he and Anatoly would've written memcached on OS X (my one contribution to the project!) first and ported to Linux to run on servers only after the proof of concept. Perhaps at some point the soup of macports/brew/fink will converge to a point where installing a recent version of emacs isn't predicated on a 4gb Xcode download. Perhaps other people don't mind the frustrating OS X window management and slow git as much as I do.

And here's an imaginary future of my own. Suppose someone came out with a distribution of Linux that only supported specifically named hardware, but supported that hardware perfectly. (That hardware could even be Apple hardware.) Suppose further that they threw away most of the Linux userland, the distribution and desktop baggage — like broken software updates, or this month's flavor of crazy new UI, or the promise of 100000 apps that all work in a janky way — and instead curated just 10 core apps to work just right.

I imagine a system where getting a project-local LAMP or node stack is a primitive operation, allowing your dev environment to closely mirror production. Perhaps you'd make the system oriented around VMs (like SmartOS) to get the delta even smaller.

Would such a developer-focused OS succeed? I know I'd use it, but I am weird. (When my designer friends ask me about the tools I use, I start with the disclaimer: I am a Morlock.) It of course wouldn't work for developing for iOS, but it would cover the web and Android. Perhaps other programmers rely too much on Mac-specific software like iTunes or TextMate, but I don't know any of those people.

But there's no money to be made in such a thing, so it will likely remain imaginary. The dark horse left out in the above discussions is ChromeOS, but as far as I can tell they are just espousing vision #1 above. (Despite working on Chrome, I have very little visibility in ChromeOS's plans; not because they are hidden, but just because I haven't looked.)

Where does that leave us hackers, then? I guess it's worth noting that Brad, cited above, gave up on OS X — "too many beachballs" — and went back to a moderately-functioning Ubuntu laptop, upon which he runs the same software as before: a browser and a terminal.

PS: what's wrong with Ubuntu? It deserves a post of its own, but it boils down to that (much like elementary) they are trying to make a computer for human beings, but they lack the focus and resources to do anything that isn't a less-reliable clone of whatever Apple does. The result is the uncanny valley effect.

[1] I qualify human need there because there is still a nice niche market for writing tools to solve computer problems. I've also intentionally overlooked the large amount of "enterprise software" yet to be written and maintained on Windows, because I just don't see any new blood picking up Windows development willingly just to get into that field.

[2] My experience of this is out of date; it's now supported by OS X, but off by default.