[Dime] Auth-Session-State (WGLC comments of draft-ietf-dime-nat-control-03)

["Frank Brockners (fbrockne)" <[e-mail]>]

Hi Jouni,

Thanks for your detailed review of draft-ietf-dime-nat-control-03. 

Wrt/ your comment on the use of " Auth-Session-State AVP", I agree that
the use of this AVP is redundant and not really useful, because I'll
have the default value in all cases, given that in DNCA both manager and
agent are stateful. In order to avoid confusion, I'd follow your
implicit suggestion to remove the AVP - and - more importantly add
additional details on how state is maintained.

Simply referring back to RFC 3588 for a description of state machines
unfortunately does not work here. Per our earlier discussions in the
working group, the nomenclature of "server" and "client" does not really
apply to the DNCA use case, hence the use of "manager" and "agent" for
the two involved Diameter entities. 

My suggestion would be that we add another section which show state
machines at agent and manager in detail (obviously reusing the format in
RFC 3588). Would that be an agreeable way forward?


Thanks & regards, Frank


> Ticket #9: http://trac.tools.ietf.org/wg/dime/trac/ticket/9
>
> During the WGLC review of draft-ietf-dime-nat-control-03 I started to
> think about the Auth-Session-State AVP and the associated semantics.
> The draft should state, even if it seems obvious, whether the state is
> maintained or not. And whether it applies to all possible messages
> exchanges. Can Auth-Session-State be on NCA or in NCR during
> QUERY_REQUEST?
> 
> How well the DNCA "view" of the session state goes along with RFC3588
> authz session state machine?