Partially-powered languages

Here's an interesting polemic about, well, everything that's not Haskell and using a declarative data model for its data structures, essentially, but especially where that concerns the Java/Ant/XSLT ecology. There are some interesting comments with corrections, but I get the author's concerns.