How Microsoft's file system caper could wrongfoot the DoJ
And annoy Larry very much indeed
A wag last year suggested to us that since Oracle was the only competition Microsoft had left, its next step would be to bundle SQL Server with the operating system.
If current briefings are correct - and Microsoft has evidently been briefing pretty heavily - then that's exactly what the company intends to do, and then some. It's a move could not only wrong foot the Beast's wealthiest antagonist, but antagonise the Department of Justice.
A fortnight ago it emerged that Blackcomb, the successor to WinXP, was to be put back with an interim release, currently dubbed 'Longhorn' (but we're not sure how much traction this codename really has) performing the evolutionary upgrade. Then on Monday IDG attributed the delay, via Gartner, to antitrust concerns about revolutionary file system changes in Blackcomb.
The Beast has a couple of goals for future Windows storage. Let's start at home.
As Jon points out, it currently maintains a scrappy mixture of file stores in Jet, Exchange folders and SQL Server itself. Exchange would benefit from running on a real database, and we suspect few would grumble from the benefits he points out.
But building SQL Server into the OS - effectively making the file system a relational database - opens a whole new ball game. Several have set out to put file system semantics over a raw native RDBMS before, rather than the other way round - and you can argue that Pick and the IBM AS/400 follow this example in its most tightly coupled form. Less tightly enmeshed are the real big iron databases which eschew native OS calls for raw I/O. (Ironically enough Oracle's 'Raw Iron' 8i database which ran on a modified version of Solaris was a lot less 'raw' than it's other offerings). That's what Blackcomb would be doing.
But what's prevented a commercial OS from delivering this is practical performance. Be Inc was the last to try, and gave up quite early on for the next best thing, a fast native file system that relies heavily on extended attributes for database-like features.
Right about now you're wondering about what kind of issues this raises for folk who need to interoperate with the Beast.
Jeremy Allison of the SAMBA team recently expressed concern that future versions of Windows would contain proprietary technology. Making NTFS essentially a device driver layer add-on drives Windows compatibility into a whole new ballpark. It doesn't have to - but it could.
It's a pretty cute idea. Cairo may have stopped being touted as soon as it was obvious that OS/2 and OpenDoc were dead, but Bill's never quite given up the idea of unifying disparate PCs as a giant distributed database. Check out our coverage of the 'Digital Dashboards' initiative a couple of springs ago here and here , or the MegaServices (as it was then) .NET announcement in May 2000 here
Bill's vision of thousands of desktop databases all unified as one distributed Windows information system pre-empted the peer-to-peer hype of last year, although the scalability issues that beset a truly distributed architecture such as Gnutella as soon as it began to be used don't really vanish for a Windows version. Never mind, you can probably imagine Bill thinking, people will buy thousands more Windows boxes as caching servers.
And simply having a database on every desktop doesn't make centralised data centres obsolete overnight: there's reliability, bandwidth and administration to think about. You don't want to go looking for an invoice only to discover it's on a distant PC that's been turned off too cool down after a particularly heavy fragging session.
But it has the almost accidental bonus of annoying Larry Ellison immensely. Isn't this what Oracle's Internet File System and Raw Iron were really about? Yes, indeed.
Finally, and by way of a bootnote, what a shame that basic file system semantics never made it into the hardware for all developers to exploit. They almost did: Object Based Storage was much fancied by the hardware manufacturers, who of course could charge more for the smarter disks than today's zero margin commodity drives. Drives would have enough knowledge to back themselves up, or replicate for example. That adventure seems to have stalled, and Microsoft has picked up the initiative with a vengeance.
"Cunning and it may even be legal," we concluded when Gates sketched out his pre- .NET vision for XMLised data stores.
We don't share Gartner's point of view that Microsoft is truly and deeply concerned about the antitrust implications. For a start, Oracle claims the lion's share of enterprise databases, and Microsoft is far from being in a monopoly position in the server market. That isn't going to change overnight. Really, there's little more reason for a delay other than that more time is needed to complete the experiment.
We're not sure if the lawyers have even clocked the magnitude of this one just yet.
MS poised to switch Windows file systems with Blackcomb
MS mounts major bid for soul of knowledge worker
How MS can hold onto the market while embracing XML
MS opens NexGen Windows Megaservices kimono
Sponsored: Benefits from the lessons learned in HPC