Minutes for DIME at IETF-96
minutes-96-dime-1
Meeting Minutes | Diameter Maintenance and Extensions (dime) WG | |
---|---|---|
Date and time | 2016-07-22 08:00 | |
Title | Minutes for DIME at IETF-96 | |
State | Active | |
Other versions | plain text | |
Last updated | 2016-07-22 |
minutes-96-dime-1
DIME IETF96 meeting Notes E2E flew through the IESG AD to push button AVP Security No feedback LS to IETF They will respond that they have an I-D (individual submission) that will be accepted as a WG document SA5 will meet at end of August On second point, they will have text from Lionel regarding the migration to 6733 Maryse Gardella (Nokia) will be at the SA5 meeting and convey the LS to the SA5 team Overload Status (Steve) Need more comments on rate control Load Control Load algorithm - is there a need to define a load distribution algorithm? (No response in room; will continue on mailing list) Another question was: do we need to define capacity or only percentage? Group Signaling (Lionel) Goal is to update the document per Steve’s feedback; get a new draft for for WGLC 4006bis (Yuval) G-S-U / U-S-U - Lionel added commentary on the fact that this is merely a clarification Zero Grant - We need to figure out if there is a protocol spec impact and clarify if the receiver of the zero G-S-U that does not understand it then we need to add a clarification that if you want to support the old method (wait for traffic, send an interrogation) then this is okay Future Proofing Enums (Maryse) Can you clarify why those are to be extended? If now we are going to extend the Equipment-Info what is the process? Lionel addressed this. In the current process of the bis if we come for a new proposal for a value do we still have time? Yes. (Stephen) It would be worth considering the privacy implications of these new types of User-Equipment-Info-Extension parameters in the draft Eventually decided to add a privacy section as part of the 4006bis work, not specific to the new IMEI type. Authors agreed to update the Security Considerations in general Another consideration is the use of RFC 7575 which will update the cypher suite (Lionel) Anything related to clarification and slight modification is fine. For any new requirements it is fine to incorporate something in a specific way that is already defined or in use is okay. However, we need to keep the changes minimal as possible. (Maryse) We stick to the refining process and direction but we is the Q-F-U-I. Will this be exclude? (Lionel) No. If it expects the old AVP then send the extra information (Q-F-U-I). (Jouni) Should we add a feature list for this? We need to consider carefully the impact on the client if not supported. We will do further analysis on the Q-F-U-I Asking for Adoption Unanimous on the adoption Authors should use the Issue Tracker Updated Design Guidelines (Yuval) - Concerned about failover which requires the Client to receive the error. (Lionel) - Will not work. From the base protocol Point of View you should view this as a delivery mechanism. Unable to Deliver would be sent back to the Client and not relying on the redirect indication. That can be used to determine failover. (Maryse) - It was her understand that the overload draft was used as an improvement for overload information. The load was introduced to anticipate any load information and the overload control is about end to end handling of the overload indication. Domain masks o (Lyle) DNS usecases are zero rating and parental control. Should be used only in non-DNSSEC deployment for privacy issues o (Stephen) also raised concern around privacy, and abuse for other usecases o (Lyle) explained usecase for SDN related bitmasks for IP and flow labels Per-flow max bitrate o (Lyle) clarified that “flows” are just IP flows (5-tuples), not replacing application level QoS. Shallow packet inspection should be used where DPI is not needed Policy Groups o (Lyle) gave the experimental background leading to the formation if the idea o (Lyle) explained the different concepts (base-names, rules bitmaps, domains, match types, membership assignment) and how the map to procedures with rules and solve inefficiencies with current approach to rule system Predicted units o (Lyle) in NFV the client should be proactive. Currently it does not have a good way to predict future load o (Lyle) the server can be a better point for predicting load based on usage profile of subscriber. And information should be sent back to client o (Lyle) current proposal based on bytes should be modified to include bits/sec, time-o-f-day infor and other things that are more useful for load prediction o (Stephen) was concerned that usage profile is usually information that is not sent back to client, and might be privacy risk (Lionel) was asking how and where the above AVPs are going to be adopted? o (Lyle) in NASREQ, by operators doing custom stuff, maybe by other groups (3GPP CT4) (Lionel) want to explore using policy groups as a mechanism for system configuration and not just subscriber provisioning o (Lyle) maybe added to group update work Warp up o (Jouni) No new charter item o (Jouni) 4006bis and group signaling could be finalized over mailing list and virtual meetings o (Steve/Lionel) security related work will require more effort, may need more face2face