> +1, why not use XSLTC today?
>
> >I have this kind of setup:
> >
> >A -> B -> C
> >
> >Stylesheet A imports B, which imports C.
> >
> >B and C are both reasonably large stylesheets.  A is generated per
> >request, is quite small and overrides specific templates in B or C
> >depending on the requirement.  I keep compiled copies of B and C in
> >memory, so when A is not needed I can just use the compiled version.
> >However, when A is needed, B and C must be parsed and built
> each time.

[snip]

> well, you have the ability to do this with XSLTC today?

In my particular case, stylesheet A is generated from another stylesheet
per request.  It's not static.  Can XSLTC help me here?  (I'm also using
2.0 now...)

> I think concerns about portability remain, and might be
> useful to discuss in the context of packaging, compiled
> stylesheets, function libraries, and so on.

Why not just complile up a version of your stylesheet for each processor
that's likely to use it?  Compiled stylesheets for me are all about
performance, not portability.

cheers
andrew