dime - Tuesday, November 11, 2014

dime

dime@jabber.ietf.org

Tuesday, November 11, 2014< ^ >

mark.jones has set the subject to: "dime wg meeting in Vancouver"

Room Configuration

Room Occupants

GMT+0
[18:57:38] stefan.winter joins the room
[19:00:06] mahoney@nostrum.com joins the room
[19:02:04] <stefan.winter> Hi, I’m Stefan, I’ll be the Jabber scribe today.
[19:02:17] <stefan.winter> Meeting started, Note Well flashed.
[19:02:25] <stefan.winter> IPR slide flashed.
[19:02:29] <stefan.winter> Agenda bashing.
[19:04:57] <stefan.winter> Status update
[19:05:08] <stefan.winter> Now at mic: Serge for Congestion and Filter attars
[19:09:11] Meetecho joins the room
[19:09:23] <stefan.winter> Diameter Group Signalling slide deck
[19:09:36] <stefan.winter> Title Slide
[19:09:43] <stefan.winter> Slide 2 (Summary 4th)
[19:10:17] <stefan.winter> Slide 3: User of Term
[19:11:13] <stefan.winter> Slide: Clarification on Permission Consider
[19:11:45] <stefan.winter> Slide: Clarification on ... (2)
[19:12:24] <stefan.winter> Slide ... (3) (Overrule a peer’s group assignment)
[19:13:03] <stefan.winter> Slide ... 4 (Delete a session Group not owned by the peer)
[19:14:07] <stefan.winter> Slide: Server to reject new group assignments
[19:19:01] <stefan.winter> Slide: client to reject ...
[19:19:11] <stefan.winter> Ben CAmbell: this needs to be supported.
[19:19:19] <stefan.winter> Seems good.
[19:19:45] <stefan.winter> Steve: having a separate flag might be better.
[19:20:11] <stefan.winter> Ben: concur. Spearate flag (explicit choice) is better than implicit one.
[19:20:27] <stefan.winter> Slide: Transaction time-out when performing Group Operations
[19:22:09] <stefan.winter> Steve: processing may slow down due to overload
[19:22:46] <stefan.winter> Maybe enough to put guidelines in, explaining how „big“ an operation should be at max
[19:23:45] <stefan.winter> Lionel: comment was rather about dependency of later result on previous big group operation
[19:24:30] <stefan.winter> Steve: text in document currently doesn’t allow early response
[19:25:05] <stefan.winter> Steve: send response only after all group operations are processed - current text
[19:25:41] <stefan.winter> Jean-Jaques: what about partial success? What then?
[19:26:07] <stefan.winter> There is text on that... fall back to single operation per request.
[19:26:30] <stefan.winter> will be discussed further on ML
[19:26:59] <stefan.winter> Lionel: two forms of feedback, early response and then completion.
[19:28:30] <stefan.winter> Slide: proxy to maintain consistency
[19:30:20] <stefan.winter> Slide: session assigned to multiple groups
[19:32:02] <stefan.winter> Steve: looks like rathole if need to determine from one session group command which other groups may be affected
[19:32:56] <stefan.winter> Need to discuss whether unaffected groups are signaled explicitly or implicitly (by omission from group command)
[19:34:19] <stefan.winter> Slide: Editorial comments
[19:35:28] <stefan.winter> Slide: next steps
[19:37:03] <stefan.winter> Slide deck: DOIC
[19:37:17] <stefan.winter> Slide: DOIC Status
[19:38:47] <stefan.winter> Slide: DOIC Status (2)
[19:39:37] <stefan.winter> Slide: open issues
[19:41:31] <stefan.winter> Slide: open issues from editor notes
[19:42:19] <stefan.winter> Slide: issue 58 options
[19:43:42] <stefan.winter> Ben Campbell: think 1 and 2 are the same thing.
[19:44:39] <stefan.winter> Ben: prefer option 1 (lets people find creative ways to fix, instead of mandating one way)
[19:45:49] <stefan.winter> Ben: solution 2 allows client to know that there are multiple sources - but why would it need to know?
[19:47:39] <stefan.winter> propose to go with bullet 1 (no change)
[19:47:50] <stefan.winter> objections? None in the room.
[19:48:14] <stefan.winter> Go bullet 1 (with clarification text)
[19:48:25] <stefan.winter> will confirm on the list
[19:48:47] <stefan.winter> Slide: agent changing OC-Supported Features AVP
[19:50:59] <stefan.winter> Ben Campbell: comment from interim: comes from an agent, is policy decision
[19:52:17] <stefan.winter> BC: extensions are not a way; don’t put normative language in this document, and then allow extensions to cancel that normative statement
[19:52:53] <stefan.winter> so don’t put normative statements around that
[19:53:20] <stefan.winter> overarching course „don’t break anything“ in your policy decisions
[19:54:17] <stefan.winter> if you do, your agent is responsible and takes the blame
[19:55:23] Catherine Dibble joins the room
[19:56:12] Dave Crocker joins the room
[19:56:24] <stefan.winter> (ongoing discussion about wording in the draft)
[19:56:43] Dave Crocker leaves the room
[19:56:53] Catherine Dibble leaves the room
[19:57:57] <stefan.winter> Ben: we have an extension which changes situation. Maybe have to add additional AVPs in OCSF.
[19:58:18] <stefan.winter> Either be completely transparent, or take responsibility for changes („you break it, you own it“)
[20:00:27] <stefan.winter> hypothetical situation: client and server support an overload algorithm C, proxy/agent in the middle does not.
[20:00:41] <stefan.winter> Agent still needs to proxy support for C so that client and server can select it.
[20:00:50] <stefan.winter> But it doesn’t understand it itself. Issue?
[20:02:16] <stefan.winter> Jean-Jacques: agent needs to ensure consistency. It’s like a gateway.
[20:02:32] <stefan.winter> Ben: yes, consistency is important
[20:03:23] <stefan.winter> all word-smith towards Jean-Jacques’ statement
[20:03:34] <stefan.winter> Slide: Remove upper limit
[20:05:24] <stefan.winter> Ben: 24h already ludicrous. Sending reports late and for large periods is pretty bad.
[20:05:45] <stefan.winter> Lionel: just not state explicit mandatory limit.
[20:06:25] <stefan.winter> Add text: long durations can have negative effects. That’s all.
[20:06:49] <stefan.winter> Ben: Default is currently 5 sec. Too short?
[20:06:58] <stefan.winter> What is the suggestion?
[20:07:20] <stefan.winter> Any number is arbitrary, like 1 Minute.
[20:07:49] <stefan.winter> 30 sec also fine.
[20:08:16] <stefan.winter> or not have a default.
[20:08:57] <stefan.winter> different proposal: make default algorithm-specific
[20:11:10] <stefan.winter> describe in AVP what timeout means, and algorithms need to define their behavior themselves (incl default)
[20:11:25] <stefan.winter> Slide: Requirement Compliance Analysis
[20:12:02] <stefan.winter> Ben: detail vs. summary section. Detail goes out of RFC, summary stays in.
[20:13:01] <stefan.winter> Benoit: also mention deadline for 3GPP was a reason; should be in the intro
[20:13:21] <stefan.winter> Benoit: detailed one should be retained „somewhere“, not in the RFC
[20:17:10] <stefan.winter> Jean-Jacques: not sure we need detailed analysis at all, so removing from RFC is fine
[20:18:20] <stefan.winter> Benoit: having a deadline in another SDO is a good reason to defend work; later revisions can happen in the IETF at any time later.
[20:19:13] <stefan.winter> Security Considerations still not finished. Work ongoing.
[20:19:52] <stefan.winter> Q: review by security expert?
[20:20:31] <stefan.winter> Ben: entire draft should be reviewed by security experts, not just the sec consideration section.
[20:21:03] <stefan.winter> Ben: because there is potential for attacks. Hopefully covered anyway, but should be looked at. Security will be in IETF LC anyway, but early review would be beneficial.
[20:21:22] jgunn joins the room
[20:23:17] <stefan.winter> More questions from mailing list in the queue. Defer to later; first get on with agenda as scheduled, come back in the end.
[20:25:28] <stefan.winter> Atle: 3GPP release 12 will be done in December. Would be good to get stable, done, document from IETF by end of 2014.
[20:27:09] <stefan.winter> Lionel: IETF process is a bit unpredictable; IETF LC and IESG review may take a while. Main body of document should be stable though; but need to keep going until publication as RFC.
[20:27:38] <stefan.winter> AVP can be allocated early.
[20:28:05] <stefan.winter> But only after doc has left the WG (completion of WGLC)
[20:28:41] <stefan.winter> Benoit: can also discuss IANA problems with people at the IANA desk!
[20:28:58] <stefan.winter> Benoit: provide feedback early, will benefit the document
[20:29:44] <stefan.winter> Jean-Jacques: what about WGLC?
[20:30:08] <stefan.winter> need new rev; once exists see if that rev is stable enough for WGLC.
[20:30:16] <stefan.winter> If so, WGLC immediately.
[20:30:45] <stefan.winter> Ben: there is an RFC and procedures for early allocations. Follow the rules.
[20:32:24] <stefan.winter> Lionel: reminder to send all comments by mail to the list
[20:32:42] <stefan.winter> Slide deck: agent overload
[20:32:48] <stefan.winter> Slide Status
[20:34:23] <stefan.winter> Slide: Overview
[20:36:43] <stefan.winter> Slide: Overview continued
[20:38:19] <stefan.winter> Slide: Peer Report Use Cases
[20:39:15] <stefan.winter> Slide: Next Steps
[20:40:45] <stefan.winter> AT&T supports this work, should move forward
[20:40:58] <stefan.winter> Ben: adopting doesn’t mean all is perfect.
[20:41:11] <stefan.winter> Ben: would argue to adopt, then work on it
[20:42:18] <stefan.winter> Who has read the draft?
[20:42:22] <stefan.winter> ca. 6 people
[20:42:51] <stefan.winter> opposed to take this as baseline? None.
[20:42:59] <stefan.winter> will confirm on ML.
[20:43:19] <jgunn> i support it
[20:44:01] <jgunn> janet gunn
[20:44:14] <stefan.winter> Slide Deck: rate control
[20:44:38] <stefan.winter> Slide: Status
[20:44:56] <stefan.winter> Slide: Overview
[20:46:26] Meetecho leaves the room
[20:47:45] Meetecho joins the room
[20:48:21] <stefan.winter> Slide: Overview continued
[20:50:22] <stefan.winter> Ben: dont disagree; just be aware that algorithm is fundamentally different than anything considered previously
[20:50:31] <stefan.winter> is also stageful
[20:50:41] <stefan.winter> strike last. *stateful*
[20:51:47] <stefan.winter> Slide: Next Steps
[20:52:59] <jgunn> have read 01. have not read 02, but support as WG draft
[20:54:06] <stefan.winter> Jean-Jacques: have to be consistent on what is to be expected from downstream node. Need to be in sync with DOIC?
[20:55:26] Michael Tuexen joins the room
[20:55:29] <stefan.winter> again ca. 6 more people have read.
[20:55:42] <stefan.winter> No other proposal on the table.
[20:55:58] <stefan.winter> Objections? None in the room. Will confirm on ML.
[20:57:24] <stefan.winter> Slide deck: architectural Considerations for Load
[20:58:19] <stefan.winter> Slide: draft is mostly questions, not answers
[20:58:24] Catherine Dibble joins the room
[20:58:30] Michael Tuexen leaves the room
[20:59:13] <stefan.winter> Slide: diff between Load and Overload
[21:00:00] Catherine Dibble leaves the room
[21:01:40] <stefan.winter> Slide: How is Load used
[21:03:53] <stefan.winter> Atle: general purpose should be considered
[21:04:26] <stefan.winter> Lionel: main use may well be load balancing or node selection
[21:07:08] <stefan.winter> Ben: there are proprietary load balancing algorithms. Would be nice to share details to make more informed decisions.
[21:07:27] <stefan.winter> Maria: load balancing is /main/ use; in order to prevent overload
[21:12:09] <stefan.winter> Slide: who uses Load
[21:14:28] <stefan.winter> Slide: who uses Load (cont)
[21:14:46] <stefan.winter> Lionel: No.
[21:17:04] <stefan.winter> Steve: No is a bit hard. Should be investigated.
[21:22:43] <stefan.winter> Slide: next steps
[21:23:14] <stefan.winter> Is writing architecture draft a worthwhile effort?
[21:23:27] <stefan.winter> Jean-Jacques: what would be the objective of an arch draft?
[21:23:38] <stefan.winter> Present framework, actors, how they interact.
[21:24:53] <stefan.winter> Atle: how appropriate is arch draft for IETF anyway? Is the discussion maybe worthwhile having in ?CT4? because 3GPP is main user of Diameter?
[21:27:27] <stefan.winter> Lionel: architecture considerations need to be in any draft anyway, whether separate or inline.
[21:28:36] <stefan.winter> Jean-Jacques: arch draft would identify decision points, and a solution draft would tackle which decision points were taken?
[21:29:17] <stefan.winter> Ben: WG adoption is far away for this draft.We don’t have a milestone for Load.
[21:29:44] <stefan.winter> Show of hand: Load in milestones? 5.
[21:29:50] <stefan.winter> Not? 0.
[21:30:01] <stefan.winter> Chairs count 7:0. Unanimous in any case.
[21:30:22] <stefan.winter> Do we need requirements beyond 7068?
[21:30:39] <stefan.winter> 3GPP might have requirements beyond that, worth documenting.
[21:31:14] <jgunn> audio died for me
[21:31:30] <stefan.winter> I think WG time is over.
[21:31:37] <stefan.winter> Maybe thats why you were cut out.
[21:32:28] <jgunn> got it bak. who knows?
[21:32:56] <stefan.winter> adjourning for lunch.
[21:33:03] <stefan.winter> Thanks, goodbye.
[21:33:11] <jgunn> thanks for being a good jabber scribe
[21:33:14] <stefan.winter> :-)
[21:33:33] stefan.winter leaves the room
[21:34:21] mahoney@nostrum.com leaves the room
[21:34:54] jgunn leaves the room
[21:34:54] Meetecho leaves the room
[22:01:49] stefan.winter joins the room
[22:17:23] stefan.winter leaves the room
[22:24:09] stefan.winter joins the room
[22:41:09] mahoney@nostrum.com joins the room
[22:43:01] stefan.winter leaves the room
[22:54:28] stefan.winter joins the room
[22:56:52] mahoney@nostrum.com leaves the room