Dynamic Metadata and Artifacts

[Rod Widdowson]

> > Do we want to issue a warning if we ever encounter an artifact query for a 'Dynamic' provider (which means regexp, template or
> > well known)?
> 
> If there are specific cases that would be known to be inappropriate (like treating the artifact source ID or location as an
entityID), that
> seems like something to block, but I don't know if that's evident, I should probably review how the old code works.

The case I'm thinking is the "standard" one.  

 In usual running
    1) We resolve http:/foo/idp dynamically
    2) The software caches it adding appropriate secondary index (from source ID or leap of faith SHA1(http:/foo/idp)
    3) The IdP responds with an artifact and SHA1(http:/foo/idp)
    4) The software gets all it needs from the cache.

In my paranoid failure case (which isn't huge artifacts or anything like that). The metadata goes out of validity or is otherwise
ejected from the cache between stages 2 and 4 and so (4) results in a cache miss and as per discussed to death the software has no
idea what to do next.  The other case is that there may be some magical profile in which an SP sees an artifact having not seen
stages (1) and (2).

Unlikely?  Yes.  Possible? Yes.

R