If you agree with this, the choice between .NET and Java is then based on (1) The background of your developers (how comfortable they are with one technology or the other) and (2) The maturity of the products and technologies within each camp.
Note that the Web services standards themselves make it easy to integrate between Java and .NET apps, and this, too, means that the decision can be pushed lower on the IT command chain, because it is no longer an enterprise-critical decision. It is just too easy (as we have seen with our own Web service orchestration technology) to get interoperability between Java and .NET Web services to require this decision to be made at the CIO level and be uniform across an organization.
The second decision factor - maturity of Web service orchestration technologies, I believe that Java has an edge *now*, pending release of the next-generation BizTalk Server from Microsoft. To get a feel for the level of maturity of the current release of BizTalk server, you just need to monitor some of the newsgroups devoted to its developers.
Hence, projects which need to be deployed prior to the end of 2003 are more likely to be successful, on-time and on-budget, if built with Java Web services and Web service orchestration technologies. For projects which don't need to begin until the middle of 2003, however, we'll have to wait and see what Microsoft has to offer and how Java Web service orchestration technologies progress between now and then.
This was first published in December 2002