That Search Thing

Okay. So I'm sitting in front of the computer having one of those "I don't know where to start" moments, and I realize I haven't been writing any background articles lately. The excruciating office pace has left me with little free time to do much more than skim the news and do the odd bit of coding, so having a cold and being housebound for a couple of days is actually a nice chance to type up some random thoughts on current trends. Here's the first one.

Search is one of those trends that everyone say coming, but that went beyond everyone's wildest predictions. And Google, of course, is the quintessential example of how a simple text field on a nearly blank page can be an earth-shattering application in itself. And with its success, everyone is trying to get on the bandwagon without considering the basics: making it simple, fast and meaningful, but never, ever complex.

For instance, the most important thing about Google was that it did things right. It dispensed with visual clutter, provided near-instant results, and rendered pages exactly the same no matter what browser I used. Unlike other search engines, it was actually designed to help people find things, not just search for them (there is a world of difference between both approaches).

Failures are quickly forgotten. It might come as a surprise to many, but Windows has had desktop search for a long time. The thing is, it has generally sucked. Microsoft had several goes at it inside Office, but all its attempts failed miserably - not only were they too complex to use, they bogged down the machine to no end - so much so that I've made a point of never again installing search aids with Office (except Lookout, which despite some issues mostly works) and disabling every variant of their indexing service I come across. It's that bad.

And now, with things like Spotlight, Beagle and Google's own Desktop Search crossing over into the desktop space, it looks like Microsoft is going to have to do something real soon.

I honestly hope they don't go the Google route, since using a local HTTP server to search for things on my own PC seems a damn poor UI approach - it should be tightly integrated into the desktop environment (like Spotlight or Quicksilver), and not require the user to use a browser at all - it's just nonsensical.

And talking about nonsense, bear in mind that people don't want to find other people's stuff on their computer - a lesson Google has recenly learned the hard way.

But there are more subtle issues: For instance, I'm not so sure the "one search box to rule them all" approach is workable - stuff like Zoë and Grokker has taught me the usefulness of targeted indexers, because you can't assume indexing e-mail is the same as indexing documents, or that the links between them are merely of statistical significance.

It is too soon to say who will win (and much, much too early to even call this a "race"), and there are, of course, no "right", predictable approaches - only ones that work, and even then I think that there will be only ones that work for specific people and not a Grand Unified Search interface.

Quicksilver and Dowser, for instance, have become indispensable to me in my daily routine, and neither of them was on the radar two years ago. So I'm not exactly holding my breath for the next big search thing.

After all, it's going to be a commodity technology real soon. The only reason it isn't already is, well... Lack of vision, I guess. The pieces have been in place for years now (there is no end of indexing toolkits, and Lucene alone has spawned more offspring than most other similar projects), but some have been too busy playing Monopoly to notice.