Is this a metadata error?

[Peter Schober]

* Cantor, Scott <cantor.2 at osu.edu> [2013-05-24 23:08]:
> Are you sure they really support it? If they're (improperly)
> expecting a single value of ePSA, maybe they didn't map the primary
> affiliation attribute at all.

Since at this point it's the IdP's problem anyway I'd probably go for
applying local logic and release the value determined as "best" (i.e.,
your ePPA value + the scope) as the only value of an ePSA attribute.
(e.g. by creating another ePSA attribute definition that has ePPA as
its dependency.).
Then the SP would only recieve the ePSA (which they probably already
can handle, from what you said before about other customers/IdPs) and
you can be sure it has nothing to do with ePPA not being processes
correctly at their end -- of course only after you've determined that
you're actually sending the right stuff (set the first logger in
logging.xml from INFO to DEBUG and either wait 10 min or restart the
context/container).

Of course by then you're knee-deep into what Scott described before,
that you'd have to have special logic for ePSA for all deviating cases
(when the subjects' affiliations to your institution really are static
and don't change depending on who's asking).
-peter