>The only downside that I see is that every ElemTemplateElement will
>have to have some boilerplate code inserted into it's execute method.
Which could impose some performance impact even when trace is not in use...
though I guess the right thing to do would be to prototype it and run
benchmarks to check whether the penalty is acceptable.
(There's a certain amount of tension between "toolability" -- debugging and
related hooks -- and performance. In the long run, I think we're expectin
that the XSLTC compiled stylesheets will become our heavily-optimized mode,
without much tool support, and interpretive Xalan will be our
debug/interactive environment. But until we reach that point, we do have to
keep a close eye on the interpreter's performance.)