Scott Cantor wrote:
>>We hadn't wanted to use wording saying the (non-schema-snippet) prose
>>literally "takes precedence", but in fact that's the only practical
>>solution when we have these added constraints. What we've got is
>>therefore confusing.
>
> It takes precedence, but only in the sense that something might be permitted
> by the schema and not permitted by the text, and is therefore not permitted.
> In the majority of cases, I think we've avoided that, but a couple still
> exist. And we probably should try and identify them and make sure they stand
> out.
>
> But the oppposite isn't true...if it's outlawed by the schema, the text in
> no way should be construed as to permit it. And the problem with English is
> that it's hard to be precise, and I just don't want to convey the impression
> that one can understand the syntax without a working knowledge of XSD.
>
> We could have used RELAX, etc., and I'm not taking a position on the
> wonderfulness or lack thereof of XSD, but it's what we used.
>
>
>>However, the normative text in this specification provides the
>>authoritative interpretation of all SAML semantics and processing. (In
>>some cases, the text deliberately defines constraints that go beyond those
>
>
>>expressed in the schema documents.)"
>
>
> I'm a little uncomfortable with the looseness of that in terms of not making
> clear that while we sometimes outlaw schema-valid XML, we never "inlaw" ;)
> schema-invalid XML.
But in practical terms, we need to give a clear signal to implementors
about what to "respect". Maybe we need to make double-sure that we
never get the prose wrong in describing the schema effects (I tried to
do a thorough job prior to cd-01 but it may have gotten out of sync
again or I might have missed something). But it's impractical to tell
people, "If the difference favors tighter constraints in the prose than
the schema, use the prose; if it favors tighter constraints in the
schema than the prose, use the schema!"
Come to think of it, I suppose we could say that the "tighter
constraint" is always in force, but then I'd worry that the meaning of
"tighter" requires interpretation...
Eve
--
Eve Maler +1 781 442 3190
Sun Microsystems cell +1 781 354 9441
Web Products, Technologies, and Standards eve.maler @ sun.com
