RE: [xml-dev] Seeking Examples of XSLT Memory Stress

["Michael Kay" <mike@saxonica.com>]

> 
> But then the semantics of unload are that those two 
> guarantees no longer 
> apply. Is that too hard to specify?
> 

Yes, it really messes up the formal semantics if calling the same function
twice with the same arguments can produce different results.

Implementations can muddle through with this kind of thing, by adding the
odd conditional to the optimizer, but it's very hard to express the
consequences formally.

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