A Healthy Dose of Windows Server 2008 Wariness

So, Windows Server 2008 Beta 3 is available. Microsoft representatives have repeatedly stated that Beta 3 is “feature complete.” Beta 3 of any Microsoft product is typically a harbinger of that product’s imminent arrival, so unless things go horribly wrong, I wouldn’t be surprised to see Microsoft release Windows 2008 to manufacturing by October or November of this year, at the latest.

And, you see, that’s why I’m worried.

I’m really looking forward to another version of Windows Server. To someone who writes, consults, and speaks about Microsoft technology, any new goodies means new grist for the mill—new fun stuff to play with, you know. But I’m actually not in any rush to get Windows 2008, given the choice between an incomplete product shipped “on time” (by whose clock, I always wonder?) and a trouble-free version shipped a bit later.

I’m a bit gun-shy about Windows 2008 because of what we just saw with Windows Vista. In its impatience to get Vista out to the market, Microsoft shipped a product that doesn’t have many huge bugs—at least not in my experience—but has a zillion little annoyances. For example, Vista lacks Vista-compatible Active Directory (AD) administrative tools (and don’t you dare suggest that I use Remote Desktop to access my domain controllers—DCs—that’s a workaround, not an answer). Or what about the small-but-essential fact that adding an external manifest to a legacy program in hopes of solving a User Account Control compatibility problem works sometimes but not always, thanks to a glitch in the disk-caching algorithm? Or the fact that I can’t get my Windows Server 2003 systems into Vista’s Network Map because there's no Link Layer Topology Discovery responder for Windows 2003? (There’s one for Windows XP, but it doesn’t load on Windows 2003.)

What will be incomplete about Windows 2008? I don't have all the answers for you. I’ve mostly worked with Windows 2008 server management, AD, name resolution, DNS, and Server Core, but at least one thing troubles me in those areas. (To be more truthful, there's one thing that I’ve got space to cover in this article. The lack of a Group Policy Management Console is another.)

Consider the mechanisms for transition from a Windows 2003-based AD environment to a completely Windows 2008-based AD implementation. Much of the functionality seems straightforward. Potentially the most significant change in AD—Read-Only Domain Controllers (RODCs) —will shake things up a bit, but not in a terribly disruptive way. Yes, Microsoft has to be careful about the way full DCs replicate with RODCs, but the company has largely solved that problem by causing RODCs to only “see” and be able to replicate with full DCs that are running Windows 2008. Windows 2008 DCs understand the special needs of RODCs, so Microsoft has cleverly sidestepped any worries about domain functional levels. Very nice.

But Windows 2008’s AD introduces a different and more significant change: SYSVOL replication. SYSVOL is a file share on every DC that holds files of crucial importance to AD, not least of which are that AD’s Group Policies. But SYSVOLs replicate not by using the robust engine that drives AD’s replication but instead by using a source of frequent trouble—the File Replication Service (FRS). Windows 2008 seeks to fix those troubles by replacing FRS with DFS-R (I’m told the acronym once meant something but no longer does)—a far sturdier engine. But how do you make the shift from SYSVOL replication with FRS to SYSVOL replication with DFS-R? The two replication methods are as different as a hand-carried message is to a FedEx package, so you can see how the transition might be sticky. I’ve been waiting to see what Windows 2008 will offer in this area. But there’s nothing in Beta 3!

That’s a bit stunning. I mean, if I were running a big AD environment such as the ones at HP, Intel, or Proctor and Gamble, I’d be worried about the lack of a testable migration tool. It would be pretty disturbing if it showed up in some interim build just six weeks before Windows 2008 debuts. I can’t imagine many beta testers would have the time to get much testing done.

On the other hand, perhaps it’s not such a big deal. After all, Microsoft didn’t discover that upgrading XP to Vista under some circumstances could blow both XP and Vista until just days before the planned Vista RTM date, and that didn’t delay Vista’s RTM by more than a couple of weeks. But I hate the suspense, you know? When clients ask me whether I strongly recommend upgrading quickly to Windows 2008, I find myself on the “Longhorns of a dilemma.”