* Douglas Teoh <
dou...@dteoh.com> [2013-06-27 07:23]:
> > Not in the way "query" is used within SAML, no. There are cases where
> > the SP needs to query the IdP for data (i.e., the other way around)
> > but you can chose not to support those SAML interactions, which makes
> > deployment much easier.
> From what I read, the IdP needs to ping the SP to get/refresh
> metadata, no?
Asking the very party you want to interact with to self-assert
security information (such as protocol end points and cryprographic
keys) does not give you any assurance or security of the fidelity of
that data.
But the answer depends on your trust model. In large multi-lateral
identity federations you have a trusted third party that vouches for
the metadata and signs it with a trusted key (how trust in that key is
bootstrapped is a different matter, let's say it happens OOB). Members
of the club then regularly pull metadata from the 3rd party, verifying
the cryptographic signature on the metadata itself.
Anyway, refreshing metadata has nothing to do with the procotocl
exchanges involved in a SAML interaction. You can keep the latter as
simple as I stated above.
Well, what about it? If you want to know more about that have a look
at the SAML specification.
Like I said, there are protocol bindings that involve the SP pulling
data directly from the IDP, mostly to prevent data passing though the
HTTP User Agent in the clear. And you don't need to support those.
Using SAML2 with encrypted SAML assertions gives you the same
protection without requring back-channel communicatons between the SP
and the IDP.
-peter