The lost promise of WinFS
Last month, Microsoft developer Quentin Clark announced in a blog
that the company would not release the second beta version of WinFS, its once-highly anticipated next generation relational file system technology. Instead, it will be rolled into the next version of SQL Server (So sensational was the news that Clark had to follow up with some clarifications
the following week).
We're disappointed. Relational file management is a radical idea, but one whose time may be anon. Today's file systems organize their contents into hierarchical trees of folders, a static and crude model for file representation. A relational model, in theory anyway, could categorize files by a wider range of characteristics.
Almost every office worker suffers from information overload, and WinFS could have tackle this problem on the desktop computer. E-mails, memo snippets, Web pages and spreadsheets and other documents could have been organized into folders with the nuance of a elegantly-written SQL query.
(Some technicalities to cover here: WinFS is not a file system per se, but rather software that would be run over Microsoft's current file system, NTFS, to provide relational capabilities. Also, Clark argues that WinFS is not entirely dead, that it will live on in SQL Server and the company's ActiveX Data Objects technology. The heated responses to Clark's post however seemed to indicate otherwise. "I don't believe those within Microsoft who made this decision realize the importance of WinFS to the future of the Windows platform," one wrote.)
While the Redmond, Wash. giant throws mad money at chasing after Google Inc. in the search-driven advertising market, and at playing catch up with Sony Corp. in the realm of console games, we can't help but to feel that it has blown an opportunity to significantly advance its own core field. Ultimately, Microsoft is in the computer platform business. So it is too bad that more of company's immense cash reserve won't go towards to tackling the hard architectural problems there, such as developing a better paradigm for file systems.
And that's only one hard problem. How about further advancing notions of trusted computing and labelled security, or even topping the excellent shell and pipelining environments that Unix now deploys? These are all the kind of advanced features that would have kept Microsoft years ahead of Linux--and kept the field progressing as a whole.
Of course, WinFS takes on a difficult problem, one that may not necessarily be solved by merely throwing more money, or assigning more engineers, to the task. 'Getting the relational engine to behave and perform like the Windows filesystem isn't a matter of a few lines of code ' it has to be done very carefully and architected right. The bars on performance, compatibility, etc. are all super high,' Clark wrote. And who knows how far the development team were from their actual target, or what state the code base was in?
But even if a relational file system approach proved to be a blind alley, at least it would been investigated. It is Microsoft's responsibility, as much as it can be any successful computer company's, to lead such deep inquiries. Oh well, now we'll have to look to EXT4
for advances in file systems.--Posted by Joab Jackson
Posted by Brad Grimes, Joab Jackson on Jul 05, 2006 at 9:39 AM