I would be interested in working out a solution where client identifier is just 
the redirection URI registered (or not), which would completely decouple client 
authentication from the rest of the flow. But that's a much bigger change.

EHL

From: Manger, James H [mailto:james.h.man...@team.telstra.com]
Sent: Wednesday, June 15, 2011 7:37 PM
To: Eran Hammer-Lahav; OAuth WG
Subject: RE: Redirection URI and Implicit grant

It seems like an authorization server receiving a request with an unregistered 
redirect_uri of https://example.org/ can tell the user:

  "Permission will be passed to your browser then onto *example.org*"

An authorization server receiving a request with a registered redirect_uri of 
https://example.net/ can tell the user:

  "Permission will be passed to your browser then onto *WidgetXYZ by Example 
Inc (example.net) [logo]*"

Where the label, company, and logo are extra details from the registration, 
hopefully after some verification by the authorization server that all these 
details are a consistent set. That verification could come from some discovery 
on the redirect_uri, or from reputation stats - perhaps even without a priori 
registration.

We might be better off ditching the client_id field and just keeping 
redirect_uri in the implicit grant. If the value is registered (or the server 
has done some discovery on it, or has some reputation stats for it) it can 
display a more user-friendly permission message.

The difference between registered and unregistered for clients without secrets 
feels like a continuum, not black and white. So separate flavours 
(client_id+state vs redirect_uri) makes less sense to me.

--
James Manger


From: Brian Eaton [mailto:bea...@google.com]
Security for the implicit grant type comes from identifying the client based on 
the redirect URI.  At client registration time, you bind client_id X to 
redirect URIs Y and Z.

If the redirect URIs use HTTPs, that gives you reasonable security.


From: oauth-boun...@ietf.org [mailto:oauth-boun...@ietf.org] On Behalf Of Eran 
Hammer-Lahav
Changes I would like to see:

The implicit grant should be split into 2 flavors (can be given different grant 
type name or just using normative language to define the restrictions), one 
with client_id and state, and one with redirect_uri only.

Registered client: the client request token response type by including its 
client identifier and optional state parameter...

Unregistered client: the client request token response type by including a 
redirection URI (no client_id or state)...

_______________________________________________
OAuth mailing list
OAuth@ietf.org
https://www.ietf.org/mailman/listinfo/oauth

Reply via email to