What’s So Hard About a Service Pack?

The end of 2004 wasn’t pretty for Microsoft. First, the company told us that Windows 2000 has no future. Then, Microsoft told us that Longhorn Server has no future. (Thankfully, the folks at Redmond eventually changed their minds.)

I guess we shouldn’t be surprised, but even though Microsoft had talked in the past about offering Win2K Service Pack 5 (SP5), the company announced late last year that it would not release such a service pack for Win2K. Instead, Microsoft is releasing an ongoing set of “we have to do this, so here it is” hot fixes for various existing and future scary bugs.

This is bad. Really bad. For a couple of reasons.

First, rolling out a Win2K Server system will remain inconvenient. The ability to slipstream service packs onto I386 machines—and thereby create new systems that have up-to-date patches already installed—has always been nice. (Anyone who’s recently tried to install Windows XP on a system directly connected to the Internet understands what I’m talking about. As the system boots for the first time, you have only a few minutes to install SP2 before someone’s unpatched system infects it with the msblast.exe worm.) XP SP2 lets you slipstream—oh, sorry, the new phrase is “integrate,” for reasons that only Microsoft grasps—hot fixes into an I386 system, letting you produce new XP images that are worm-proof from the moment they boot up. But apparently that functionality will never be possible in Win2K Server. Bummer.

Second, the notion of support is an even larger concern. Microsoft has tried to settle on an OS lifecycle plan over the past few years. The company’s aim seems to be, in essence, “Don’t get really comfortable with any of our software, because we’re going to force you off of it in a short time so that you’ll have to buy our newer version, no matter how minor or expensive the upgrade is.” This mindset offends me beyond words.

I know I’ve used this analogy in print before, but its common sense remains valid: In 1986, I purchased a Honda Civic. I don’t drive it anymore, but a member of my family does, and if I want support from Honda, I can get it. Sure, the support comes at a price, but no one at Honda would say, “You know, once that thing’s spark plugs wear out, you won’t be able to buy replacements from us. Instead, you can upgrade to our 2005 models.”

Of course, Honda doesn’t have to do that. Microsoft didn’t have to do that, either—until the past few years. Nobody in Redmond had to put a gun to the heads of Windows NT 3.51 users and say, “You must upgrade.” That wasn’t necessary, because NT 4.0 was just plain better. Folks have seen the advantages of Win2K over NT 4.0, but it’s taken almost 5 years for us to get to the point at which some people are ready and willing to move to Win2K—only to learn that Microsoft soon plans to shove them off that platform as well.

So Microsoft has released a bunch of hot fixes into a “rollup” but won’t package the items into a service pack. Would it be that difficult to test these hot fixes sufficiently to create a service pack? Clearly, in the minds of some in Redmond, such an effort is unnecessary. Is it truly reasonable to brand Microsoft customers who are happy with Win2K as second-class citizens? Or is it just a sales tactic to get people to upgrade?

Once upon a time, automobiles represented a brand-new technology that was so unreliable that you had to upgrade every few years. But gradually the automobile industry matured, and today car owners can—if they choose—reasonably expect to drive a car for 10, 15, perhaps 20 years. And those owners can obtain whatever level of support they desire and are willing to pay for from their dealer.

Is the software business much different? Both are technological businesses, and both offer diminishing levels of technological progress, like other technological fields. When the field matures, we can expect large software vendors—particularly those reporting an 87 percent profit rate, as Microsoft did last year on Windows—to offer full and undiminished support over longer periods of time.

Can we see any proof of my suggestion that the rate of change in Windows’ improvements is slowing down? Sure—look at Microsoft’s decision to put off WinFS, as well as its recent decision to also remove WinFS from Longhorn Server. Which means we won’t see WinFS in a Microsoft product until at least—are you ready?—the year 2011.

Microsoft keeps stretching the length of time between server releases (which should yield programmer time for maintenance of current releases) while simultaneously offering less and less in the way of maintenance releases (i.e., service packs). I repeat: How difficult is it to produce a service pack when all the hot fixes already exist? With Longhorn slipping and Windows Server 2003 Release 2 (R2) looking less and less exciting all the time, perhaps people are crunching the numbers and wondering why they need to upgrade from Win2K. Is Microsoft giving them a reason?