RE: Scope of uniqueness of generate-id()

> Until today I was thinking that xslt functions have no side 
> effects (provided extensions are not considered). 
> It has turned out that this is not correct.
> That changes the picture (at least for me).
> Many possible optimizations are not safe.

Yes, you will find this discussed in the XQuery formal semantics, I believe.

It's a very limited kind of side-effect - but it does have to be taken into
account by an optimizer.

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