LocalMetadata SAMLtest transition issue

[Nate Klingenstein]

> Couldn't an mtime of the inode that's newer than maxCacheDuration for
> a given metadata provider trigger a re-parsing, without having to
> essentially turn off cashing completely?
> (I.e., inodes in the thesourceDirectory could be stat'ed quickly and
> therefore much more frequently?)

I'm going to drip noob here, but I think I understood what you said.  When a new metadata file with an entityID that matches an existing one is uploaded to SAMLtest, the timestamp is indeed refreshed.  The one I was toying around with:

-rw-r--r-- 1 apache apache  2832 Mar 19 16:44 e965d5fcae878af978a3c3babeb6ccdb1a96e1c4.xml

This has no effect on whether the cached version or the new version is used, so I think the caching mechanism and timers are entirely internal to the IdP, but I would be in favor of such a change personally as it would allow me to remove those attributes from the MetadataProvider.