The Libre Web Application Stack (A Code Story)

[This was originally published at autonomo.us– comments over there.]

[Disclaimer: since my last post at autonomo.us, I have become an employee of the Mozilla Corporation. I don’t feel this has tainted my views, but feel free to weigh that information as part of your analysis of this article. Relatedly, I do not speak for that employer when writing here; my words and ideas here are my own.]

tl;dr version: We shouldn’t throw the baby out with the bathwater. The bathwater is the ethical problems with software hosted on someone else’s server; the baby is the strength of the libre web application development stack- which may actually be the best platform for building autonomy-preserving user-oriented software.

Put yourself in 1995. I’m going to tell the you of 1995 that in 2010, there will be a software platform with the following properties:

• The widget toolkit for this platform will have two independent, competitive, Libre implementations; one of those implementations will be the most widely used consumer-facing piece of free software ever, and the other one will have the support of the second and third largest software companies on earth.
• Tens of millions of dollars a year in engineering time will go into improving those free implementations of the standard; in fact, Microsoft will spend lots of money and PR time saying ‘we’ve caught up to the libre implementations.’
• This widget toolkit will have no vendor gatekeepers or tolls; in fact, all the biggest software vendors on earth will have attempted to disclaim patent claims against it, making it not only no-cost but also perhaps the safest free software toolkit available from an IP perspective. New additions to the toolkit will be free by default; there will be some exceptions to this but those exceptions will be bitterly contested, sometimes with huge proprietary software companies weighing in on the side of free implementations.
• The major implementations of the widget toolkits will be extensible using a simple scripting language, making them much easier for users to customize than most (non-emacs) toolkits.
• Free software written for this platform can be made trivially available not just to users of free operating systems but also to people on other platforms, potentially helping grow the community of free software developers and users.
• This widget toolkit will be available on every computing platform on earth; not just PCs but also many phones and soon on TVs. If your hardware supports this widget toolkit, it immediately has enough applications that it is considered commercially viable, leading to an unprecedented blossoming of operating systems platforms (unlike the situation in 1995, where new platforms have no apps and so find it impossible to compete with Win3.1/95.)
• The logic and data storage systems backing this widget framework will also be varied- there will be proprietary implementations, but there will be dozens if not hundreds of free frameworks as well, in essentially every programming language you can imagine. It is taken for granted that these frameworks run on Linux first. Among many others, the White House will use such a free framework to deliver software.
• At the same time, because of elegant design, it is easy to swap out backends, so if you want to write backends in different ways, you can do that; in fact, people are taking advantage of this every day to write completely new frameworks with a variety of different properties.
• A function called ‘view source’ is viewed as a key competitive differentiator in this platform; it is actually hard to close the source describing widget layout and behavior (though easy to close the backend.)
• Millions of high school kids are taught (at least part of) this framework; millions more will have taught it to themselves using view source!
• This platform will have been so successful that virtually every single first-world computer user will use it (in some way) every day.

At this point, 1995-you says ‘This sounds too good to be true. There must be a catch.’¹

There are two major catches, the second a consequence of the first.²

The first catch is that the default data storage architecture is distributed, and the default licensing is effectively permissive, making it expected (though not mandatory) that users are separated from both the code they run and their data. In other words, developers who use this stack are perhaps more free than they’ve ever been in the history of computing (and not surprisingly they’ve adopted it in droves)- but users often have even less control than they had when they were using traditional proprietary desktop applications.

The you from 1995 would probably think that freedom-lovers would have reacted to this unfreedom by rewriting the thin layer of proprietary code sandwiched between gigantic gobs of free code, and/or by working to make the platform more amenable to local or distributed use. You might expect that they would even have embraced and extended the platform to make it even better for freedom.

Instead, the second catch is that fans of freedom have largely thrown the baby out with the bathwater, ignoring (or at best failing to embrace) this rich, free platform. Instead, in a story straight out of the Innovator’s Dilemma, they’ve continued to focus on traditional widget toolkits and interfaces which lack all the benefits I’ve just listed out.

It should be obvious by this point that the platform I’m talking about is HTML and the many, many application frameworks that can be used to generate it. This platform- what I’ll call the libre web application stack- is not perfect, but it has the potential to be hugely freedom-enhancing, and free software advocates should carefully consider it when thinking about developing their next app, or when planning to extend traditional software platforms to make them more free. Just because most web apps are not free does not mean the stack itself should be disregarded.

1. This is true whether 1995-you is a free software fan or not; if you’re a free software fan, you’re particularly excited, but if you’re a die-hard silicon valley capitalist you should also be pretty excited by the potential of this platform. [↩]
2. There are plenty of minor catches too; performance isn’t great; the widget toolkit does not have the vast variety of widgets that more mature toolkits have. But these have been improving consistently for years, and (more importantly) they’ve been improving much faster than traditional toolkits have improved. [↩]