> > Let's make sure that "eval" (or similar) is included in the 
> next XSLT 
> > revision.
> 
>   This would be useful indeed.  Maybe a good way to improve 
> our chances to see it in a further revision is to write a 
> formal specification and implement it in several processors.  
> You are more than welcome to join EXPath 
> <http://www.expath.org/> and propose something.
> 
>   The key point, I think, is to define it precisely regarding 
> the static and dynamic contexts.

We're working on a specification for XSLT 2.1. The key insight was that you
can give much more control over the static and dynamic context if you make
it an instruction rather than a function; for example you can use child
<with-param> elements to bind variables referenced in the dynamic
expression.

Regards,

Michael Kay
http://www.saxonica.com/
http://twitter.com/michaelhkay