Doing the impossible with Longhorn

It’s funny how circumstances can change your perception of what’s possible. A few months ago, key Microsoft architects were telling me that it would be impossible to decouple the Avalon presentation subsystem from the Longhorn OS. Now they’re huddling in conference rooms trying to figure out how to do just that. It makes me wonder what else might turn out to be possible after all.

In our July 19 cover story on Longhorn, I evaluated the three “pillars” — Indigo, WinFS (Windows File System), and Avalon — in terms of benefit and lock-in. The Indigo communication system has the most attractive benefit/lock-in ratio. It’s firmly rooted in Web services standards, and it isn’t tied to the Longhorn OS. So last Friday’s announcement that Indigo will be made available in 2006 — for Windows XP and Windows Server 2003 — came as no surprise.

Demoting WinFS from a Longhorn pillar to an optional component that will be slipstreamed in later was a bit more surprising. But that’s what can happen when you try to do something genuinely innovative, as I believe WinFS is. My interview with Quentin Clark , director of program management for WinFS at Microsoft (see also the extended version on my blog), explores the subject in depth. I think WinFS embodies a correct and historically inevitable strategy that can, with appropriate mappings to XML standards, yield an attractive benefit/lock-in ratio. I never expected such an ambitious synthesis of object, relational, and XML data management disciplines to be fully baked anytime soon, so the recent announcement of a WinFS delay is hardly shocking.

Here’s an interesting footnote to the WinFS news: According to John Montgomery, director of product management for the developer division at Microsoft, good old full-text search will play a larger role in Longhorn. Empowering us to find and organize our stuff was, after all, one of the major goals of the project. The early rhetoric discounted full-text search in favor of the highly structured WinFS approach and suggested it would be impossible to deliver the desired benefits any other way. Now architects are huddling in conference rooms trying to figure out how to do the impossible. Of course, Apple had already previewed a similar strategy for the forthcoming Tiger version of OS X. If Steve Jobs can demonstrate Spotlight in 2004 and if Apple can ship it in 2005, Microsoft ought to be able to match that by 2006.

The problem pillar, for me, is Avalon. Initially, the benefit/lock-in ratio looked especially bad for enterprise customers. The “presentation experiences” envisioned by the Avalon architects, who aim to create a seamless blend of document, user interface, and media elements, played to the consumer desktop more than to the enterprise desktop. And businesses that deploy consumer-facing software, as nearly all do, were presented with a brutal choice: To develop for Avalon meant sacrificing not only Web reach but also Windows reach. The advice to businesses was to suck it up and invest in a software stack that few existing PCs are even able to run. When push came to shove, they wouldn’t take that leap of faith. So now, Avalon will also show up on Windows XP and Windows Server 2003.

It remains an article of faith in Redmond that the Web platform has run out of gas. Do you think there might be some unacknowledged possibilities there as well? I do.