The long road to Longhorn

Developers and development project leads have some hard labor ahead when Longhorn establishes a wholly new Microsoft platform. It’s true that Longhorn keeps slipping, slipping, slipping into future, but you should treat these delays not with sighs of frustration, but sighs of relief. You want this ship behind schedule because you would never have made it to the dock on time.

Even if you decide you don’t care about Longhorn, be aware of the likelihood that the OS you’re using now will be dead by Longhorn’s arrival. Microsoft has pruned Windows NT 4.0 from its list of supported servers. What little life Windows 2000 Server has left in it will be stomped out by the Longhorn server (whatever it is ultimately called). It’s Windows Server 2003’s world now; don’t fight it. Longhorn’s only living relative on the client side will be Windows XP (which is being remade in Service Pack 2).

Windows developers haven’t had to wrestle with multiple hardware platforms since Windows NT 4.0. Well, set your clocks back 10 years, because we’re getting back on the platform-go-round. The ride will start with Visual Studio 2005, which will target 32-bit Pentium, Xeon, and clone hardware, along with the 64-bit AMD 64, Itanium, and Intel’s x86-64 extended-memory architecture (its answer to Opteron). The little ones, Pocket PC and Smartphone, will toddle up behind. Developers would be wise to maintain reference machines for each platform so they can test-target freshly compiled applications. If that’s too complicated or costly, the next best thing would be to study warnings from the compilers and from VS 2005’s static analysis tools, which will do their best to flag code that isn’t portable.

Finally and probably most importantly, developers should move everything to XML as fast as they can. Persistence of application and object states, data-stream formatting, document representation, and GUI structure should be in XML by Longhorn time. No one can predict Microsoft’s specific schema for each of these, but the wonderful thing about XML is that you can always get there from here.

Clearly, developers will have some latitude in defining their own migration paths as Longhorn rumbles nigh. But it’s absolutely certain that nailing code to a particular version of Windows, a particular hardware platform, or a particular way of structuring and presenting GUIs and data will negate many of the advantages of new Windows tools and platforms.