Hi Rob,
I read the "cached request" posting as well. I didn't pick up enough
useful information from that thread to discuss. Are they suggesting
that customers generate assertions valid for 6-12 months? That does
not seem wise. If a "validation window" needs to be set to 6-12
months to support a "cached request" approach, I believe it's at
Google. For example, if they are checking the InResponseTo attribute
to correlate a response to a request, then they (Google) would need to
recognize the cached request ID as valid for an extended period of
time... of course, they can't be checking the InResponseTo attribute,
otherwise IdP-initiated wouldn't work to begin with.
I re-tested IdP-initiated SSO to verify accounts in both my primary
and non-primary domains. As long as I'm posting to the ACS in the
primary domain, and fully-qualifying the Subject's NameID (i.e.
us...@foo.edu or
us...@bar.foo.edu), it is successful. Like you, I am
hesitant to pursue this option unless Google will provide some more
formal statement or documentation. Searches have led me repeatedly to
salesforce.com, which seems to support and document IdP-initiated SAML
SSO. That doesn't really have anything to do with anything, except
that it gave me hope Google might support the same approach.
On Jun 19, 8:08 pm, Robert Norris <
r...@eatenbyagrue.org> wrote:
> I know from experience that secondary domains do not work with IdP-initiated
> SSO, only the primary domain. We never got around to confirming with Google
> whether or not IdP-initiated SSO was supposed to work or was just a leftover
> from something. We guessed that since it wasn't documented and wasn't
> exactly a common practice it would be best if we switched SP-initiated.
>
> Something we successfully tested but didn't implement was making a request
> on the server side to the service URL, extracting the SAML request from the
> redirect response that came back, filling out the detail and signing it and
> then giving that back to the user. From the user perspective it looks
> exactly like IdP-initiated SSO (ie a single redirect). This approach might
> work for you.
>
> (fyi, this is similar conceptually to the "cached request" approach
> described in this old forum thread:
http://www.google.com/support/forum/p/apps-apis/thread?tid=0dee17907b...)
>
> Cheers,
> Rob.
> > The deprecated Reference Implementation<
http://code.google.com/googleapps/domain/sso/saml_reference_implement...> for
> > SAML-based SSO to Google Apps still works fine with IdP Initiated SSO. I
> > put up a Proof of Concept to do the same, and it also works fine for our own
> > domain. Google is accepting our signed SAML response with a valid
> > RelayState. I opened a support case to inquire whether or not Google would
> > *continue* to accept these unsolicited responses, and I was redirected to
> > this support forum. Can a Google employee comment further or commit to an
> > up or down for continued support of IdP initiated SAML SSO?
>
> > I would also like to describe my scenario and current understanding, and
> > perhaps there's an alternative approach someone could suggest. Our users
> > will be in multiple domains within our overall Google Apps environment.
> > Most users will have a single account, but some users may have an account
> > in more than one domain. More than one account would result in more than
> > one link to access Google Apps. My curent understanding for SAML with
> > multiple domains is:
>
> > 1. I need to post the SAML response to the ACS for my primary domain,
> > and only to my primary domain
> > 2. I need to fully qualify the username in the SAML response, for
> > example
u...@foo.edu or
u...@bar.foo.edu