The Web Application Layer

This post is an attempt to ask "what next?" in the world of contemporary application development. I'm disturbed by the conveyance of applications in this format. This is not news to popular readers, but, rather than complain extensively about the state of the contemporary technology, I think it would be more productive to muse on possible improvements and some of the underlying structural concerns in this space

In No Particular Order...

Today, we routinely design and implement user interfaces in HTML and JavaScript these days. I'm not convinced that HTML, or any XML based format is really all that good for conveying well formatted structured text, much less pixel perfect graphic design and application interface.

Lightweight text markups like Markdown, reStructuredText (for all it's warts), and Textile are human readable, provide structure, and convey text well. Furthermore, it's very possible to efficiently translate them into very high quality output formats, including XML formats and LaTeX.

One of the driving forces behind the convergence on "web-technologies" is that JavaScrpt/HTML/CSS are all thought to be "cross platform" technologies. It doesn't matter if you're running on a Mac, or a PC, or UNIX system, if it has a web browser it'll run there. The web application movement realizes the "write once run everywhere," notion that Sun attempted with Java in the 90s (except, that Java never really worked for that.) Except that every browser implements JavaScript/HTML/CSS in a different way which means, that it's really "write once and tweak it to death so that IE/Firefox/Webkit don't break." There are some things (like jQuery; HTML5) that make this better, but the browser market is dirty and browser makers will never be incentivized to comply with the standards. [1]

RESTful APIs [2] are, I think, leading to more desktop applications. Or at least making them more possible. It used to be very much the case that if you wanted web-connected data you had to go to a website. Now, if you want data from the Internet, in most cases it can be gotten in an easy to process formant (i.e. YAML/JSON) and then folded into a desktop application. In addition to "rounded corner power" and "social media," the biggest impact of "Web 2.0" has been the increasing awareness and interest the API[^quality] in the more general public.

Adobe AIR is a wonderful idea. Even smaller lightweight devices like smartphones and netbooks are so powerful that it doesn't make a lot of sense to have them operate as such "dumb" clients. Conventional web development has developers cobble together server applications that put together content and then chuck it off to the client for rendering. With APIs (see above) it doesn't make a lot of sense to leave all the heavy lifting on the server. Adobe understood this with AIR, the problem with AIR? It's wildly proprietary, applications look out of place on every platform, and performance is miserable relative to "real" native applications. Its a great idea, and I'm terribly interested to see what comes next in this space.

I'm grumpy about HTML 5 because I remain unconvinced that web standards are really a viable way of regulating sane design and development practices. It also seems too likely that HTML 5 solves the problems we were having 2 years ago, rather than the problems we'll have over the next 10 years. Also, I think this world needs a hell of a lot less XML, in any form.

What are your thoughts?

[^qualit]y]:I'm not sure there is any singular aspect of the whole "Web 2.0 thing" that is unequivocally bad or good. I think on the whole web design is better now than it used to be, but "rounded corner power" at this point all looks the same, and it's really difficult to achieve in a clean technological sort of way. And the web has always been social; so while that's not new, it's nice that the web has caught on even if the whole "social networking silo" phenomena is less than desirable. The same thing goes for RESTful APIs: it's great that data is more accessible, but it sucks that APIs can be so proprietary and I'm not convinced that HTTP is the "right" or "best" protocol for this technology. But these things happen.

[1]You may think that I'm simply being pessimistic, and you might say that IE 7 and 8 are a huge step in the right direction and I think that this might be true, but the only reason to create and maintain a browser (to my mind) are: masochism, to get people to use your search engine, and to be able to implement special proprietary (and non-standard) features. The competitive advantage comes from the unique enhancements that a given browser is able to offer over the other browsers in the market. For a while (e.g. 1999-2007?) the more standards compliant a browser was the better pages looked in it. I'm not sure that will continue to hold true.
[2]This is a simplification, but lets think of this as the obligatory API that all web-services provide today: from twitter to flickr, to YouTube, and beyond. These allow programmers to connect to the service using the HTTP protocol
comments powered by Disqus