I finally got off my backside and added an RSS feed.

DogBiscuit looks crap in Internet Explorer. Unbelievably, today is the first time I've browsed this way with TheBorg's OneTrueBrowser (I'm typically a Konqueror/Firebird user). Under IE, the entry titles gradually creep to the left! By about 2-pixels/entry, from what I can tell. Weird. Better do some CSS debugging.

Digging into it, it looks like I've hit the Magik Creeping Text bug. Removing the border-left on the blog-entries fixed it.

Phew ... lucky no-one reads this stuff!

A new feature in RubLog allows "global" URIs for articles to be constructed using WikiWords. Such URIs use just the base-name of the article - not the whole category-path - and so make links less likely to break if the category hierarchy is re-organised.

PragDave makes a valid point: categorisation is kind of artificial and arbitrary, and may be prone to change. Actually, even with only a handful of articles here on DogBiscuit, I've already been tempted to re-categorise :-)

The RubLog feature works by allowing links using just the last part (the "basename") of the article ID. If the same basename is used in multiple categories, this may result in collisions ... RubLog solves the problem by presenting all the matching articles. One problem here is that the user has no way of knowing which article is the "right one", ie. the intended destination of the link.

Categories as meta-data?

Another approach might be to capture categories in article meta-data, rather than using the file-system structure, e.g.

    Subject: Dealing with Recategorisation
    Category: /Tech/Projects/Eyaw
    
    A new feature ...

This way, articles can be filed into a directory structure that stays fairly constant (e.g. directory per day?), and so can be used to form "permanent" links. Categories are laid "on top", and it could even allow for an articles to be cross-filed in multiple categories. Holy dueling taxonomies!

On the downside, it becomes harder to see the category hierarchy, and modifying it potentially involves editing a whole bunch of files. Bother.

So, I'm messing around trying to craft a weblog CGI-script using Ruby. I've been very much influenced by blosxom, and by Dave Thomas's work on RubLog ... but I needed a good excuse to play with Ruby, so have started to hack together my own.

Basic features I'm after include:

• Wiki-esque markup
• Easy hyperlinking
• A simple, file-based data-store
• Easy navigation

But rather perversely, I'm not prepared to trust file mtimes to specify posting dates. Instead, I'm going to pull meta-information from the file "header". Sample input looks something like this:

    Subject: entry title
    Date: 13 Apr 2003, 11pm

    Here's the content ...