rousse> I guess you're thinking of making A and B member of Child substitution group, rousse> and requiring a sequence of 1 to 2 Child element in parent ?

Yes, although maxOccurs was unclear for me (might be nubmer of possible characters).

rousse> It won't solve rousse> the problem of two A or two B, as any other constraint considering child rousse> element separatly.

Ah. I didn't read Gregor's description as excluding multiplication of children; is it trouble if the Parent has multiple children, Gregor?

-- James