Longhorn server sent out to pasture

Beginning with Windows NT 3.1, the initial version, this paired versioning was a central strategy for NT, and was followed through on every version through Windows 2000.

But it's more than 10 years later and Microsoft customers just don't do what they're told the way they used to. For a variety of reasons, it doesn't make sense for Microsoft even to try to sync up releases of client and server operating systems anymore. This is good news for customers, because the pairing certainly doesn't serve their interests. One can only hope that the cancellation of Longhorn server is also a sign that major version releases will be more spread out than they have been in recent years.

How many of you planned and implemented simultaneous deployments of Windows NT 4 Server and Workstation, or of Windows 2000 Server and Professional? Anybody? I didn't think so. (Actually, there may have been a few enterprises that planned dual-Windows 2000 deployments in order to really buy into Active Directory--but there weren't many of them and they're probably still having nightmares about it.) Microsoft has a consistent history of designing these products so that you don't have to upgrade to both, much as you might want to in order to get all the new capabilities. The sane way to go about it is to move either your servers or your clients and then move to the other. There are arguments for either order.

Everyone knew it already, but even Microsoft concedes now that customers would prefer that Microsoft slow their new version rollouts. "Another major Windows server release in the Longhorn timeframe does not meet the needs of most of our customers," said a Microsoft spokeswoman. The delay "is a response to what our customers are asking for."

The rapidity of releases and the pairing of client and server releases are different issues, but they're both manifestations of release schedules being driven by the wrong incentives. The upcoming release of Windows .Net Server is a good example of a product with a set of new features that isn't very compelling. True, there are some important new features, such as IIS6, but overall it doesn't seem like a major version and calling it ".Net Server" is a serious overstatement; there's nothing about it that makes it more of a server for .Net than Windows 2000 Server with the .Net Framework installed. There are some features that make it a sort of Windows XP Server, such as group policies that manage XP-specific client features, but if we still used version numbers we'd call this a .1 release, or maybe a .5.

As for Microsoft's interests, they have many reasons to end client/server release pairing. Back in the days of Windows NT 3.x, the actual difference in the code between the two versions was miniscule. They were essentially the same programs with different settings, as was reported in PC Week when I was there. (The reported parlor trick of turning an NT Workstation into an NT Server never impressed me. One could probably as easily hack a 10-user version of Netware into a 100-user version; you're paying for a product with different capabilities.)

There were even worse abuses of naming confusion in the past. How many people got confused by the OS/2-PS/2 thing and thought you needed one to run the other, when nothing of the sort was true? I'm sure there has been some element of this with Windows NT clients and servers.

But in more recent history, especially beginning with Windows 2000, the code bases have begun to diverge substantially. Windows 2000 Server has Active Directory, Terminal Services, DFS, and dozens of other substantial features not found in the client. This makes it much harder to coordinate the releases without either slack in the schedule of one project at the end of the other's life cycle, or the release of one product a little too early because the other needs to go to market.

Add to that the increased diversity of versions on both the client and server sides and getting all the ducks in a row just isn't worth it. When Blackcomb comes along, will they simultaneously release the Tablet and Media Center client versions, not to mention the Datacenter server version? And by then they'll probably have even more new variations. Indeed, we may see schedule diversity increase among these variations, to the point where it will mean less to call them "Windows XP" versions, or whatever it will be called then.

And in the bigger picture, decoupling the products puts them on schedules related to their own merits. This is good for customers because assuming you want the product at all, you want to be able to evaluate it for its own merits and you can have some added confidence that it will be completely cooked, even if it exists more to fill a hole in a schedule than to bring really new features.

Have you ever upgraded your servers and clients at the same time--and survived to tell about it? TalkBack below or e-mail us.