The Butt Ugly Weblog

Tech brief

Don picked up my posting on yawning, and also seems to be involved in a discussion about Wiki APIs.

First, a short explanation on the tech behind this weblog: This is an instance of JSPWiki, where each entry is a separate WikiPage. The Main page aggregates then all of the pages which have a certain signature in their name onto the front page, producing the weblog you see right now. This allows cool stuff like doing collection pages, such as Ropecon2003, or EGC2003, where I just insert a string like [{WeblogPlugin startDate='310103' days='31'}] to get all of the entries from January 2003, for example.

JSPWiki (and thus, this weblog) support the XML-RPC Wiki API, as well as the MetaWeblogAPI. The support for the latter is not complete, but it is quite enough so that I can moblog from my cell phone. What happens is that I basically send an email to a predefined address, where a custom script takes the email, parses it and posts it to the weblog using the said MetaWeblog API. The script is smart enough to take the attachments from the email and modify the entry so that they are properly inserted in the text. Of course, I have to write my entry using WikiMarkup, but trust me, it is far easier to do on a cell phone than HTML!

Yes, all Wikis use different markup rules - but actually, that's not really a big problem. So far, the biggest use of the WikiRPCInterface has been adding functionality into JSPWiki itself - a machine halfway across the world uses it to check up what has happened on the JSPWiki main site, and sends email to everyone who has subscribed to it. This is a highly important thing - I did not have to implement that by myself; I just implemented an API and someone else wrote it (thanks Mahlen!). This is really no different from writing a proper, plugin-based architecture for a Wiki, except that the external developer is no longer bound to Java, or Perl, or Python, or PHP, or whatever language the lead developer of the wiki decided to use. In essence, it allows you to write simple plugins and do batch processing on a Wiki without having to understand its internal implementation!

In fact, defining a standard markup language for Wikis (such as WikiML) should work well with this kind of an XML-RPC/SOAP API approach: just add another method that gets the text in WikiML instead of rendered HTML/WikiMarkup. It does not matter what markup the wikis themselves use.

I also believe that Wikis will converge slowly. It seems to me that MoinMoin style markup is winning, and with the stuff like Radeox from the SnipSnap folks, you can separate the rendering of the markup from the WikiEngine itself, making it largely irrelevant what markup people use, since you can just plug in the markup you want to use.

However, there will always be differences in markup. You see, doing stuff like complex TWiki markup just is out of the reach of the smaller players, and since making a basic wiki is really easy, people are not going to bother to use the same markup. They might adopt some common practices, but some wikis need stuff like syntax highlighting, and others will need direct drawing on page, etc... In the end, the one and only solution is to get rid of the markup altogether, and make a WYSIWYG wiki - sort of a distributed version of MS Word.

Phew. There is already a Wiki API standard, but so far not too many seem to have adopted it. I would very much like to see discussion what is needed, what is too much, and how to make it better.