Hi Dolph,

On 21 Feb 2014, at 03:05, Dolph Mathews <[email protected]> wrote:

> 
> On Thu, Feb 20, 2014 at 4:18 AM, Marco Fargetta <[email protected]> 
> wrote:
> Dear all,
> 
> I am interested to the integration of SAML with keystone and I am analysing
> the following blueprint and its implementation:
> 
> https://blueprints.launchpad.net/keystone/+spec/saml-id
> 
> https://review.openstack.org/#/c/71353/
> 
> 
> Looking at the code there is something I cannot undertand. In the code it 
> seems you
> will use apache httpd with mod_shib (or other alternatives) to parse saml 
> assertion
> and the code inside keystone will read only the values extrapolated by the 
> front-end server.
> 
> That's correct (for icehouse development, at least).
>  
> 
> If this is the case, it is not clear to me why you need to register the IdPs, 
> with its certificate,
> in keystone using the new federation API. You can filter the IdP in the 
> server so why do you need this extra list?
> What is the use of the IdP list and the certificate?
> 
> This reflects our original design, and it has evolved a bit from the original 
> design to be a bit more simplified. With the additional dependency on 
> mod_shib / mod_mellon, we are no longer implementing the certificates API, 
> but we do still need the IdP API. The IdP API specifically allows us to track 
> the source of an identity, and apply the correct authorization mapping 
> (producing the project- and domain-based role assignments that OpenStack is 
> accostomed to to) to the federated attributes coming from mod_shib / 
> mod_mellon. The benefit is that federated identities from one source can have 
> a different level of authorization than those identities from a different 
> source, even if they (theoretically) had the exact same SAML assertions.
>  
> 

The idea to distinguish the IdPs make sense but for SAML I think it is better 
to work with the attributes only. If the SAML is the same you may work at 
mod_shib level to
create attributes containing the value you want so it is easier to create a new 
attribute in the SP having the authorisation level. In this way you can define 
authorisation rules
using only assertion instead of mixing assertion with IdP names, with …… . I do 
not know if you can exploit the same approach with OpenId and/or other 
authentication framework.

Nevertheless, it seems that I am late for icehouse so I will better analyse 
what you provide now and think what could be done in Juno.

Cheers,
Marco


> Is still this implementation open to discussion or the design is frozen for 
> the icehouse release?
> 
> It is certainly still open to discussion (and the implementation open to 
> review!), but we're past feature proposal freeze; anything that would require 
> new development (beyond what is already in review) will have to wait a few 
> weeks for Juno.
>  
> 
> Thanks in advance,
> Marco
> 
> _______________________________________________
> OpenStack-dev mailing list
> [email protected]
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
> 
> 
> _______________________________________________
> OpenStack-dev mailing list
> [email protected]
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev

Attachment: smime.p7s
Description: S/MIME cryptographic signature

_______________________________________________
OpenStack-dev mailing list
[email protected]
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev

Reply via email to