Windows Server 2003's bits are modular, not MIA

Microsoft has come up with an innovative explanation for the number of pieces that are going to have to be retro-fitted to Windows Server 2003, after it launches with them missing next month. This is apparently a modular approach to software development which will equip the company better to respond to the speed of update delivery that characterises open source.

An amazing number of birds are slain with this particular stone. Microsoft has been wrestling with the problem of a: How it delivers the Longhorn updates for Server 2003 without b: Dumping customers with the service pack from hell or c: Rolling out a 'Longhorn Server' version of the OS when it's supposed to be being good by d: Not forcing its business customers to upgrade servers every two minutes or thereabouts. At the same time e: It needs to get the updates out there and in use before its product roadmaps recede somewhere in the general direction of 2020. Which is what would happen if it only added new functionality with new server releases, not in-between, and only made those server releases at intervals sufficiently wide for businesses not to scream in pain and anger.

Confused? So is Brian Valentine, who seems only to open his mouth to change feet these days. Valentine first said there'd be no Longhorn Server last year, but intimated that the delivery mechanism for updates might be something of a problem. Then last week he said there might be a server version after all. Then there mightn't. Or there might be a Limited Edition version. And now with the new "modular" approach there probably won't be a Longhorn Server after all, just lots of stuff to stick into 2003.

On the other hand, perhaps we shouldn't altogether rule out the possibility of features being rolled up into a single, handy distribution at some point a year or two down the line. But that wouldn't be a new OS, no sir, that would be a handy roll-up.

News of the modular development comes, sort of, in a Joe Wilcox CNET piece. And we say "sort of" here because Joe doesn't attribute much of the meat to Microsoft itself - it's analysts' comment that gives the story its legs. The Microsoft input relevant to modularity comes from Bob O'Brien, who appears to have spent the past week walking behind Brian Valentine with a dart gun:

"These (non-shipping) capabilities we said from the beginning we would deliver post-delivery of the server platform. The things we're going to release later simply are going to add value to the product."

O'Brien also describes the additional "components" as things that "add more capabilities - power tools - for our customers," and the announced ones at least won't be charged for.

But it's not clear to what extent Microsoft itself is presenting this as a counter to open source, or how much of this can be ascribed to analysts' reading of the changes. That said, by introducing a delivery route that is not a whole new OS and at the same time does not involve rolling new features into service packs (it's promised businesses several times it won't do this), Microsoft will be running an update schedule which will inevitably be compared to open source ones.

Is this a solution? Well, it'll involve businesses having to make a lot of changes in order to keep their servers up to speed, and if you consider that one of the points of Red Hat's "Advanced" line of product is that it's intended not to present businesses with changes at such high velocity, you might consider the possibility that Redmond is dealing with the wrong problem here. It's come up with a formula that will allow it to shove out stuff as and when it's ready, but it hasn't answered the question: "Jeez, why d'you keep dumping me with all this stuff?" The reverse, actually, we'd say.

On which subject, Mary-Jo at Microsoft Watch has taken a shot at a roadmap of some of this stuff. Figuring this lot out then having to figure it out all over again six months down the line is a thankless task, so we're deeply grateful. You can find it here. ®