Evaluable criteria

[Brent Putman]

On 6/2/15 9:20 AM, Marvin Addison wrote:
>
>
>
> I'm still trying to digest this whole thread, but bottom line for me
> afaict is that there needs to be some changes to resolve/resolveSingle
> in the MetadataResolver type hierarchy to support my CAS use case of
> looking up an EntityDescriptor by endpoint URL.

Yes.  But the operative phrase there is "looking up".  I would say
that's correct, vis-a-vis the terminology I've been using.  Meaning it's
not a predicate.  It can not be the case (at least by default) that the
resolver accepts *only* predicate criteria which it then applies to an
iteration of all its potentially 1000's of EntityDescriptors.  Maybe as
an option you can turn on, with the appropriate caveats re: knowing the
size of the metadata you have. Certainly would not work IRL for
something like InCommon or UK federation metadata.

And that's for batch resolvers.  When you are calling this for real, you
don't know what kinds of resolvers you have in the chain.  For a dynamic
resolver, a predicate criterion simply doesn't work.  You have to have a
lookup criterion that exposes some data which the resolver can turn into
the appropriate parameter(s) for the call the metadata responder.

So in summary, supporting arbitrary predicate criteria to a metadata
resolver (or any resolver) is something we could add for niche cases,
but doesn't really help with either your CAS case, or the SAML artifact
case.  For those we need (in batch) secondary indexes.  That's going to
be a bit of conceptual work.  I'm not even actually sure how to support
non-equality-based index lookups.  It's sort of analagous I guess to
whatever RDBMS's do to efficiently support substring matching on indexed
columns on e.g. a "where name like '%Joe%'. Or what search software like
Lucene does to support full text searching.  I get the sense that
there's a whole bunch of computer science in there that I don't know. 
Hopefully there are some libraries to help us out... (maybe even Lucene?).

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://shibboleth.net/pipermail/dev/attachments/20150602/fe8e3007/attachment.html>