Applying Ourselves to Metadata Models

If operating systems can't evolve unified storage, developers must find their own way.

In Robert Heinleins classic 1966 novel, "The Moon is a Harsh Mistress," a computer technician has to explain to a newly awakened conscious machine that some jokes are only funny once. The first time, he tells the computer, youre a wit; the second time, a half-wit. "Geometric progression?," the computer inquires. "Or worse," he grimly replies.

This brings us to the subject of Microsofts much-touted WinFS, for which the great need was clearly described by Microsoft SQL Server Director Tom Rizzo in March—but which the company now says it will not deliver until after 2007.

Whats the connection between the WinFS delay and Heinleins notion of "funny once"? Well, I managed to get at least a little bit of humor from the initial promise of WinFS, observing in my column of ten months ago that we had heard all of this before. (Speaking of science-fiction writers, Jerry Pournelles blog entry from the Microsoft Professional Developer Conference records our discussion of that blast from the past.)

In that November 2003 newsletter, I was actually able to recycle eight entire paragraphs from a column written in June of 1996, describing Microsofts indefinite postponement of the object-oriented "Cairo" storage initiative that had previously been promised—in all seriousness, supported by extensive white papers—for what was then still being called Windows NT 5.0.

It seemed almost funny at the time. But only almost, and now the joke is wearing thin at a hypergeometric rate. We really did need a vastly improved storage model back in the mid-1990s; were not going to get it on Windows until the late 00s. At least, not as an integral part of the dominant fat-client operating system.

In 2005, we can actually expect that Apple will offer its "Spotlight" solution, comprising three different integrated search capabilities—for files, metadata and indexed data—that will give users streamlined, location-independent access to information without the kind of foundational change that Microsoft is now entering its second decade of attempting to deliver. Developers on the Mac OS X platform will also get new leverage in using these core capabilities, based on Apples plans for an integrated programming facility merging objects, scripts, and conventional compiled-code modules.

Metadata search facilities on Mac OS X might seem to be of interest only to Mac-specific developers, but Apple also announced this summer that its Rendezvous discovery protocol for "zero-configuration" networking is being extended to Windows and open-source platforms as well. This suggests a vision for developers in which the best available metadata engine, whether implemented by Apple or by someone else, may be transparently available to users of any client that can find a network connection—either wired or, increasingly, wireless—to that engine.

That vision reinforces the comments of Sun Microsystems President and Chief Operating Officer Jonathan Schwartz: He had not yet assumed those responsibilities when he said last November that Microsoft was at risk of being so late that Longhorns target problem would no longer exist when it finally shipped "after their next slip in 2007." Schwartz, of course, was arguing for a Java-based model, but he still made the right point when he talked about the need "to build a network platform—not a single destination platform." Whether or not you believe that PCs will even be relevant in 2008, it does seem as if developers will want to write applications that find their way to any number of other devices as well.

And applications value to users, increasingly, will hinge on success in giving users anytime/anywhere access based on metadata—"Show me the files that relate to the phone call that I got from my boss while I was driving to the client meeting last Tuesday," for example. Apple is saying, in effect, "We wont build that into our foundation, but well give you the tools to make it look that way." Sun is saying, in effect, "We dont know who will do it best, but well give you a model that lets you use it wherever it is."

Neither of those promises sounds as good as the promise of Longhorn, in 2006, with integrated metadata-based storage. That promise, as viewed by Microsoft Application Platforms Solutions Specialist Randy Holloway last October (before he joined the company earlier this year), led Holloway to say (and led many others to agree) that "in terms of potential to boost user productivity, WinFS has to be the number one feature in this product ... WinFS will be one of the core features that really makes Windows a better computing experience for the average user. Simply put, WinFS will enable you to do things with Windows that you cannot do today or that you cant do today in any reasonable amount of time."

But either Sun right now, or Apple (most likely) next year, offers a developer a path toward building solutions in (ahem) a reasonable amount of time. The ever-slipping advent of WinFS, the Unified Storage Model Formerly Known As Cairo, is a joke thats just not funny any more.

Tell me what would make your users smile at


Check out eWEEK.coms Application Development Center for the latest news, reviews and analysis in programming environments and developer tools.