Missing the Inflection Point
We'll know a few years from now. I hope he is wrong, since I plan on continuing my application development on WinForms and shipping with the Whidbey runtime. I do see a straight-forward migration path to Avalon as long as I separate interface and code; even then, there is WinForm-Avalon interoperability as well as common managed infrastructure. I also see WinForms as a version of the API that will run on legacy machines. I think customers will still want rich client applications, especially if it adopts some of the web advantages as well.
It could be that our views are heavily dictated by our prior investments. Joel is heavily invested in old Microsoft (Win32/COM), and that investment appears to (but not necessarily) be vaporizing under the influx of a succession of APIs, WinForms and Avalon. He's at the dying end of an S-curve. As such, he rails against the deprecation of the Win32 and the new initiatives of Longhorn. At square one again, he's furious at Microsoft and just reevaluating all the alternatives.
I, on the other hand, see .NET swooping along at the beginning of a long new S-curve. When I left Microsoft four years ago, I was sick of the Win32 mess and was seriously considering non-Microsoft technologies like Delphi and Java. Then, I bought into Microsoft's story about a rosy managed future and invested a few years learning the technology. And, honestly, I think that the introduction of .NET was the inflection point.