Page:Aaron Swartz s A Programmable Web An Unfinished Work.pdf/21

                                           9

From billboards, buses, and boxes they peer out at us like alien symbols (except hopefully less threatening): URLs are everywhere. Most obviously, they appear at the top of the browser window while people are using your website, but they also appear in a myriad of other contexts: in the status bar when someone has the mouse over a link, in search results, in emails, in blogs, read over the phone, written down on napkins, listed in bibliographies, printed on business cards and t-shirts and mousepads and bumper stickers. They’re versatile little symbols.

Furthermore, URLs have to last. Those t-shirts and links and blogs will not disappear simply because you decided to reorganize your server, or move to a different operating system, or got promoted and replaced by a subordinate (or voted out of ofﬁce). They will last for years and years to come, so your URLs must last with them.

Moreover, URLs do not just exist as isolated entities (like “http://example.org/lunch/bacon.html”). They combine to form patterns (“bacon.html”, “lettuce.html”, “tomato.html”). And each of these patterns ﬁnds its place in a larger path of interaction (“/”, “/lunch/”, “/lunch/bacon.html”).

Because of all this, URLs cannot be some side-effect or afterthought, as many seem to wish. Designing URLs is the most important part of building a web application and has to be done ﬁrst. Encoded in their design are a whole series of implicit assumptions about what your site is about, how it is structured, and how it should be used; all important and largely-unavoidable questions.

Unfortunately, many tools for building web applications try to hide such questions from you, preventing you from designing URLs at all. Instead, they present their own interface to the programmer, from which they generate URLs by making up random numbers or storing cookies or worse. (Nowadays, with Ajax and Flash,