RE: [security-services] AI-19 re: Versioning in protocol messages andassertions

>Actually, perhaps another question needs to be addressed before the
>specific questions below.  That is, do we need this version info in our
>schema at all?  The SAML protocol and assertion versions can be directly
>determined from the name of the schema files we define.  Is this adequate?

This was discussed a little among a group of interested parties that included me, and my opinion was that we ought to address the
issue of the schema version distinctly from the protocol and message versions, so as not to require changing the schema namespace if
the schema doesn't change in a backward compatible way.

IOW, if SAML 1.1 is backward compatible schema-wise, then the namespace should still stay 1.0, even if the MinorVersion went to 1 in
messages sent across the wire. So SAML 1.1 might use schema version 1.0.

The implication of *that* is that without the version attributes in the schema, there would not be a way to indicate any sort of
change to the processing rules intended, unless the namespace were changed.

I think namespace change is a pretty drastic and expensive thing to do to applications, and shouldn't necessarily be needed for
every new release of the spec.

-- Scott