RE: Motivations for restricting the "all" group

> I think the feeling now is that there are other, non-state 
> machine ways to implement validation of constructs such as 
> xs:all with unrestricted cardinality and so the restrictions 
> are relaxed.

Yes, it's very easy to implement the validator with counters. The hardest
part in implementing the 1.1 spec is the rules for valid restrictions, for
example the fact that sequence(a, b, a?, b, a?) is a valid restriction of
all(a{1,5}, b{1,2}). And it's a little frustrating having to implement such
features when you know no-one is ever likely to write such a thing!

Regards,

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