Metadata support: EntitiesDescriptor/@Name handling

[Brent Putman]

On 8/16/13 3:05 PM, Cantor, Scott wrote:
>
> We would, but I think your thought was that we would port those existing
> mechanisms to rely on some other input,

Yes, the input being either say an EntityGroup "object metadata"
instance(s) (just wraps a string), or EntityAttributes.

I was forgetting that we already have generalized functors for evaling
the latter.  But if we did turn @Name -> EntityAttributes then, for
backwards compat with v2 configs, I imagine we'd still have to support
the old EntityGroup functors, and they'd just be re-implemented so that
they look for a specific entity attribute with a specific name. So just
a specialization of the EntityAttributes functors.

>  and I'm not sure that's practical,
> *because* the likely alternative way to do groups is already represented
> in the code by the EntityAttributes mechanism.

Off-hand EntityAttributes did seem to me to be a more natural way to do
this, as long as there's either a standard attribute name/profile, or we
at least make up something local, in a Shib namespace.

> So essentially, if we wanted to keep the group stuff alive, we probably
> would need to standardize a custom entity attribute to use. And that seems
> kind of confusing/messy to me, and since there really is little or no use
> of EntitiesDescriptor beyond very simple one-level cases, I'm not sure
> it's worth turning them into EntityAttributes.

I don't really have a strong opinion on perpetuating the use of entity
groups or not.  But I think in v3 we can't abandon the notion entirely
b/c there's too many people using that, and we'd need a bit of lead time
to "deprecate" that usage.

>
> When I say "deprecate" I don't really mean the code, I mean the approach.

Yeah, I get it.  But unless it's literally going to be unsupported in
v3, we have to come up with something.

>
> Yes. I mean that we would consider any use of those functors a red flag. A
> SHOULD NOT if you will.


Ok.  Would you mind summarizing though what you think we ought to do
right now, in terms of the metadata resolver code?  I can't quite
parse/combine all of the above, and I'm not sure if you're suggesting we
should (for v3) represent @Name as an extensions entity attribute, or as
object metadata.