views:

463

answers:

1

Hi

After reading the MSDN article (http://msdn.microsoft.com/en-us/magazine/2009.01.genevests.aspx) on implementing a Custom STS using the Microsoft Geneva Framework I am a bit puzzled about one of the scenarios covered there. This scenario is shown in figure 13 of the above referenced article.

My questions are around how does the RP initiate the call to the RP-STS in order to pass on the already obtained claims from the IP-STS? How does the desired method DeleteOrder() get turned into a Claim Request for the Action claim from the RP-STS which responds with the Action claim with a value Delete which authorizes the call? I also think the figure is slightly incorrect in that the interaction between the RP-STS and the Policy Engine should have the Claims and arrows the other way around.

I can see the structure but it's not clear what is provided by Geneva/WCF and what has to be done in code inside the RP, which would seem a bit odd since we could not protect the DeleteOrder method with a PrincipalPermission demand for the Delete "permission" but would have to demand a Role first then obtain the fine-grained claim of the Delete Action after that point.

If I have missed the point (since I cannot find this case covered easily on the Web), then apologies!

Thanks in advance.

+1  A: 

I asked the same question on the Geneva Forum at http://social.msdn.microsoft.com/Forums/en/Geneva/thread/d10c556c-1ec5-409d-8c25-bee2933e85ea?prof=required and got this reply:

Hi Dokie,

I wondered this same thing when I read that article. As I've pondered how such a scenario would be implemented, I've come up w/ two ideas:

  1. The RP is actually configured to require claims from the RP-STS; the RP-STS requires a security token from the IP-STS. As a result, when the subject requests a resource of the RP, it bounces him to the RP-STS who bounces him to the IP-STS. After authenticating there, he is bounced back to the RP-STS, the identity-centric claims are transformed into those necessary to make an authorization decision and returned to the RP.

  2. The RP is configured to have an interceptor (e.g., an AuthorizationPolicy if it's a WCF service) that grabs the call, sees the identity-centric claims, creates an RST (using the WSTrustClient), passes it to the RP-STS, that service expands the claims into new one that are returned to the RP, and the RP makes an authorization decision.

I've never implemented this, but, if I were going to, I would explore those two ideas further.

HTH!


Regards,

Travis Spencer

So I will try option 2 first and see if that works out then formulate an answer here.

Dokie
Actually option 2 is flawed in that such an AuthorizationPolicy does not provide access to the action/method context so I have moved on now to consider what is provided in Geneva Beta 2, which looks a bit more promising in this regard of Authorization claims.
Dokie