Dave Winer wants Google to write a program that indexes his hard drive.

This is way outside Google's sphere of excellence. Google's search technology is designed to index things that are part of a hyperlinked environment (the core being their Page Rank™ algorithms), and designed to run on big, dedicated servers where the most important factor is speed of search through a massive database.

This can sort of translate to intranets - you have the hyperlinked environment, and the search tool would run on a server. It doesn't translate at all to local hard drives, where you would need to optimise the index in the other direction (space over speed), and none of the documents you are searching through are linked to anything else. The only leverage Google would get from their existing product is the trusted brand-name, and a few routines to index PDFs and Word documents.

On top of this, Google would be putting themselves firmly in the sights of Microsoft, who have made no secret of wanting to replace their filesystem with a database (an idea that is at least five years overdue, and final proof that the Linux guys don't innovate - or we'd have done this ages ago). Presumably once Windows has SQL Server in the background, full-text indexing is only a matter of an add-in module. That would really put Google in the same boat as Netscape.