Remember when every new software release was like Christmas?

In the early days of the PC, when Moore’s Law had a purpose other than compensating for bloated code, new releases contained wonderful new features. We looked forward to learning them and putting them to work. They made our work lives better, easier … and yes, more fun, too.

When was the last time you looked at a new software release that way?

Today’s topic is Windows 2000. It has, I’m sure, dozens of new features that will make our work lives better, easier, and more fun. So many, in fact, that the combined efforts of Santa Claus, the Easter Bunny, and an army of tooth fairies will be needed to carry them all.

But I promised myself I’d avoid sarcasm in this column. The question is whether you should upgrade, not whether the new release rivals the marvels of old. The answer, unsurprisingly, is, “It depends.” Look elsewhere for boldness.

There’s only one circumstance in which you should bring Windows 2000 into your company immediately, and that’s if you’re about to start coding software that will run on Windows in production. Whatever the hazards of early adoption, the wasted effort of converting a brand new application is worse.

Otherwise, you’re in no hurry. Wait at least three months for a “slipstream” release or service pack, keep an eye on BugNet, and let others pave the upgrade road for you. The rest of this column is written from this future vantage point.

Now … what are your options? One is to ignore Windows 2000 for now. Y2K remediation efforts have delayed a lot of IT projects already, and Windows 2000 is an infrastructure change, not a value-adding application. If it offers nothing you want, put your resources elsewhere.

You could migrate just your desktops, or make just your new desktops Windows 2000. The desktop is the most certain choice for Windows 2000. It should be non-disruptive there, and people I respect tell me it’s more stable than its predecessors. Test it for compatibility with your applications and use the desktop to get started.

I take claims of stability with a grain of salt, though, because so far I haven’t heard Microsoft utter the five words it’s avoided throughout the history of Windows: “We will respect our DLLs.” Given Microsoft’s sponsorship of the software licensing regulation, the Uniform Computer Information Transactions Act (UCITA), this promise would be empty anyway I suppose, but it would be a start. The specifications of every DLL that ships with Windows 2000 should be published and fixed … but of course, the same measure would have made every other version of Windows far more stable, and Microsoft has never taken it before, either.

Still, reports of improved stability may cause you to consider upgrading your Windows 4.x servers as well. Here, though, the question is more complicated, because Windows 2000 is for Microsoft what Netware 4.0 was for Novell. Let me explain.

Netware 4.0 was a radical departure from earlier versions, largely because of NDS. A move from Netware 3.0 looked more like a conversion than an upgrade, and many IS departments saw a conversion to NT, newly shipping at the time, as no more disruptive.

Migrating your servers from Windows 4.x to Windows 2000 will be a thoroughly disruptive experience, as documented elsewhere in this issue. So much so that you may find it easier converting to NDS for your global directory; Netware 5.0 or the Linux/Samba combo for file and print services; Netware 5.0, Linux, or a commercial UNIX when it comes to your intranet and your database servers (unless, of course, you use Active Server Pages or SQL*Server).

It’s only in your application servers, with all of their custom code, that a conversion is probably far more painful than an upgrade.

Windows 2000 offers you a wonderful opportunity. By being so disruptive, it allows you to take a fresh look at a big chunk of your technical architecture (specifically, the middle tier of the infrastructure layer).

Who knows … there may even be a place in there for Windows 2000.