> Of course something like this is needed, no question about that. You
> mention the relation to <AttributeConsumingService> at the outset, but
> then you neglect to use <RequestedAttribute> in any of your proposed
> solutions. Why not carry <RequestedAttribute> elements in the authn
> request (in some way), with exactly the same semantics as in metadata?
Yeah, I'd have to say that's always the solution I envisioned (with or
without a new wrapper element).
I'm not in favor of tunnelling. The existing query element is designed as a
top level PDU and it's awkward to try to reuse it internally.
Boxcarring is obviously an option, but should be at the binding level and
that's a much bigger change that I would rather avoid if there's a simpler
way.
> The other thing I'll comment on is the proposed use of
> Attribute/AttributeValue to convey non-attribute information to the
> IdP. This too diverges from current usage, and I would suggest we
> find some other way to carry this information.
+1
-- Scott
