On Fri, May 13, 2005 at 07:49:58AM +1000, Dimitre Novatchev wrote:
> > The debate with memo-function would be about whether it actually has any
> > semantics, or is merely a performance hint. Could a conformant processor
> > ignore it? What is the effect on a "creative" function, one that constructs
> > new nodes each time it is called?
> 
> xsl:function -s with side effects should not have been allowed -- in
> the first place.
> 
> So it is not only a nice wish to think about memoisation, but probably
> a pressing need to clean up the spec from functions with side 
> effects.
> 
> Or be prepared for all kinds of a nasty surprise following the fact
> that the value of
>    
>        my:f($x) is my:f($x)
> 
> is generally not guaranteed to be true()

Is there some other mechanism which would be used to,
for example, return a random number, other than a function?

> 
> 
> 
> Cheers,
> Dimitre Novatchev.