Some ideas for the next version of Basie.

Yesterday I went to see a presentation by Kosta Zabashta about the work he did on Basie. The presentation went fairly well, but while listening to it, I had a few ideas that I thought I should write down.

First, since the configurations for Exim and Postfix are so complicated and so different, it would be nice if we had a program that could figure out which one you’re running, and generate the lines to add to your config files to get it all set up.

Second, I think we probably don’t need to add that many boolean or set operators to the search functionality. I seem to remember reading a paper a while ago where Google said that people only used 2 of the advanced search features, "s to make a phrase, and OR to choose between two different things. Sure, some people used all the wacky operators, but it was a vanishingly small percentage.

Third, do we update the search index when items change, or are we constantly rebuilding it? I would have thought that doing the dynamic updates would have run into many of the same issues as the Events app, but maybe there was a smarter way to do it that I’m overlooking.

Fourth, and finally, I wonder if there’s a way to use metadata about the objects we’re searching to influence rankings. To take the Google approach, add the number of mail messages/commit logs/wiki pages/tickets that refer to an item as a factor in determining the ranking. Or, push more recent things higher in the search results, since we know when things were added. Do we actually want to do either of those? Maybe, maybe not, but we’ve got a lot of data about each item, and it seems like using it might be, well, maybe more interesting than useful. :)

So, there are my thoughts on stuff we might want to do for the next version of Basie. I suppose the next step will be to get Greg to link to this, and get other people commenting. (And now that I’ve written it all out, I suppose it might have been cleverer to post this to the Basie blog. Ah well. Live and learn.)