Paul Andrews: Microsoft to give PCs a little Google

Longhorn promises to give Microsoft a powerful edge in search and all that it signifies for advertising, subscriptions and user allegiance. But it may open up an opportunity for Google as well. Leveraging the lingua franca of XML in Longhorn, Google could develop a "bot" or utility that, downloaded onto your PC, would provide Google's features for your hard disk.

To me, this begs the question: how do you rank the hits? The most important feature of Google is its accuracy. Accuracy comes from page rank, which comes from a complex soup of update frequency, incoming links, and a potpourri of other factors.

Now shift gears to your hard drive. To be sure, there's a lot of great stuff in there. In a day, I could probably write code to index all of the text and pdf files on a chosen hard drive. In a week, I could probably write something to index MS Office files (piggybacking on Microsoft's Indexing service built into IIS). It would be brute force and ugly, but it would work.

Searching the data soup would constitute another relatively trivial task, at least for implied and searches, where all search terms had to be there, similarly for implied or searches, where one of the search terms had to be there.

Now that I think about it, Zope has a cross-platform catalog/index/search function built in. I'd read through that code and "leverage" it.

Sorting presents a major problem. Most recently updated or created first? Most instances of the search terms? Hyperlinking does not exist in the world of your documents, so it can't help. You could build its intelligence by pre-ranking some of your files and folders as more important than others, but we don't want to go there needlessly.

Historical Note: When I worked at MSFT (97-99), we all knew about this issue. Too bad they went for an Apple solution (i.e., waiting for perfection), rather than doing it quick-and-dirty.   

1:21:41 PM  comment []