CAS attribute naming

[Cantor, Scott]

On 6/13/17, 4:36 AM, "users on behalf of Rod Widdowson" <users-bounces at shibboleth.net on behalf of rdw at steadingsoftware.com> wrote:

> This ties in with the whole "want to source 'attributes on the wire directly' from DataConnectors" issue.

Yes.

> - Entirely exbedding encoding from the attribute definition (and thus collapsing the somewhat weird distinction between 
> "attributes from a dataconnector" and "attributes from a definition" - although that then gets into some very subtle dependency
> issues.

Yeah. I'd like to be able to avoid disrupting the resolver configuration more than need be with any changes, but I also think once you make encoding separate, the distinction just stops making much sense. We haven't shipped the new Dependency changes just yet, so there's time to rethink some of it.

> - Having some sort of "per protocol (or profile?)" encoder with some automagical mechanism to generate the name.  You could
> consider CAS to be a degenerate case of that.  I falter at the "automagic" bit though.  Velocity? 

I'm not thinking magic, I'm just thinking some kind of RADIUS-like dictionary approach where we have protocol dictionaries mapping from the local names. I just think having it all in one place, either in the resolver or outside it, is better than the inside out approach for some use cases.
 
-- Scott