DIME WG IETF 85; 120 min meeting notes Agenda and WG update: ======================== Presenters: 10 min - Jouni Korhonen, Chair - Lionel Morand, Chair Note taker - Jean Mahoney No Jabber scribe Note well, IPR notice, and agenda presented. Thanks for work on 3588bis, now 6733. Total 6 RFCs published since IETF#84. ACTION: draft-ietf-dime-group-signaling needs additional editors. Marco has indicated that he's willing, but everyone is encouraged to contribute for the completion of this charter item. RFC4005-bis UTF8String issue brought up during Last Call: ACTION: Lionel will ask for a review from the precis wg, but won't hold up work waiting for feedback. Peter St. Andre was identified as a helpful resource. If 6733's content on domain names is sufficient, then maybe it could be referenced. WG Documents: ================================ draft-ietf-dime-realm-based-redirect ___ 5 min Presenter: Tom Taylor Tom said that Ruibing Hao's email has bounced. Tom presented changes to the draft. ACTION: Tom to send email to Avi to see if the remaining tracker item can be cleared. After that the document is ready to move forward. draft-ietf-dime-overload-reqs _________ 15 min Presenter: Eric McMurry Tom Taylor offered to review the document. Eric presented changes since Vancouver. Remaining actions - editorial, precision, clarity ACTION: Jouni asked the working group to provide comments. ACTION: Glen recommended putting out -01 already if there are changes. Eric will do that. Diameter overload control: =================== draft-roach-dime-overload-ctrl ________ 30 min Presenter: Adam Roach Note: an IPR declaration has been made against this document. However, it was not visible on the Dime tools page during the presentation. These proposals (both overload control proposals) aren't competing could be deployed in different scenarios or merged. Slide - Overload-Metric AVP ACTION: Clarify that this is the overload control that you want to apply. Slide - Load AVP ACTION: Clarify that the information here is useful for load balancing and is not for overload control. Slide - Why can't we fail non-controlled connections? ACTION: Adam to ensure that the M-bit is not set on the AVPs. Slide - Open Issues ACTION: Open issues are called out in the draft, wg should look at them and comment on the mailing list. draft-korhonen-dime-ovl ________________20 min Presenter: Jouni Korhonen Slide - Modes of operations Eric - when something reports overload, the receiver has maintain state to deal with it. How does that work in stateless mode? Jouni - Talking about session state. You must have overload state or it won't work. Slide - Issues Under Consideration: Jouni may drop the State Maintaining Mode. Do we need more scope ? Like sessions or groups ? Ben - in large-scale implementations, there are pools and when you have scopes higher than connection level, the nodes will have to coordinate. That's expensive. That's why Adam paired his mandatory scopes down. ========================= Discussion Lionel - Not a race between 2 solutions, but best solution for the overload control. Pieces could be combined. Consider impact on protocol, signaling, implementations. We've received IPR disclosure on Adam's solution. Ben - Under the assumption that we need both mechanisms, it would be nice if we could have common data model and set of AVPs. The scopes may be different for hop-by-hop and interconnect. Question on the app mechanism - if we have an agent intermediary - doesn't that [??] Can I find out if it is something on the back side that can support it? ^^^^^^^^^^^^^^^^^^^^^ Jouni - rely on existing routing. The end2end example, MME - HSS with a bunch of boxes in between. If there are proxies that do not support the overload application in between, then you are screwed. Eric - That's true for either mechanism if the proxy doesn't support them. I support a common data model. Lionel (as individual) - The mechanisms may need different mandatory supports. There will be Diameter nodes in the networks, in roaming networks, that won't support either of these mechanisms. Need a minimal set of info to support. (MME/HSS interface) support different kinds of deployments. The client may know the mechanism. Or only the proxy at the edge of the network may know the mechanism. Negotiation between edge clients. There will be router in the middle. Support end2end and end2middle. ACTION: Define what the intermediary should/can do. ACTION: Discuss on the mailing list a minimum set of info and how to deal with it. Taking into consideration the scenarios discussed. Find a solution quickly. ACTION: Describe the data model which would be common for all overload control colutions, then figure out the transport. Diameter e2e security: ======================= draft-korhonen-dime-e2e-security _______ 15 min Presenter: Jouni Korhonen Glen: Isn't there a milestone for a problem statement and requirements; not for the solution. Jouni: Yes, you are right. ACTION: need a problem statement and a requirements document for end-2-end security. Show of hands - Lionel, Hannes, Jouni, Glen, Tom will work on it.