- Friday, November 8, 2013

aqm@jabber.ietf.org

Friday, November 8, 2013< ^ >

Room Configuration

Room Occupants

GMT+0
[20:17:09] wesley.m.eddy joins the room
[20:35:08] g.white joins the room
[20:37:15] ROADRUNNER joins the room
[20:38:54] ROADRUNNER leaves the room
[20:40:27] jgunn joins the room
[20:42:04] Mirja joins the room
[20:42:18] Gabriel Zigelboim joins the room
[20:50:22] <g.white> it’s more complicated than just setting different thresholds for drop vs. mark. CoDel schedules drops at specific future times. It isn’t clear how to graft ECN onto it.
[20:54:28] Wes George joins the room
[20:54:33] <g.white> comment was regarding the different criteria....
[20:54:37] Wes George leaves the room
[20:55:34] <g.white> if drop and mark are considered equivalent, then it is easy (as Dave T indicated)
[20:59:11] yrz joins the room
[21:01:18] <Mirja> does 6 belong in this document?
[21:01:49] <Gabriel Zigelboim> drop and mark should not be considered equivalent.
[21:02:05] <Gabriel Zigelboim> you can mark before you have to drop
[21:02:19] <Gabriel Zigelboim> no?
[21:02:25] <Mirja> this is the new requirement they want to add
[21:02:26] <wesley.m.eddy> Mirja: gorry suggests to read the document to understand 6
[21:02:44] <wesley.m.eddy> there is more text explaining it in the document
[21:02:48] <Mirja> k
[21:02:49] <Gabriel Zigelboim> who wants to add?
[21:02:54] <wesley.m.eddy> he isn't talking near the mic though
[21:03:26] <Mirja> @gabriel: there was already the discussion to add this recommendation to the document
[21:04:08] <Gabriel Zigelboim> @Mirja: not contesting it-- trying to understand
[21:04:30] <Mirja> or at least that network devices should be able to implement merk and drop rate differently
[21:04:42] <Mirja> merk=mark
[21:04:55] <Gabriel Zigelboim> who is speaking now? I'm on the audio BCast and his voice is staurated. Can someone ask him to move away a little from the mic?
[21:05:03] <wesley.m.eddy> Jim Rosskind
[21:05:12] <wesley.m.eddy> he is about a foot away from it
[21:05:13] <g.white> audio not saturated on my end...
[21:05:20] <Mirja> mine is good
[21:06:00] <Gabriel Zigelboim> not completely sat, but not real good on my end over the IETFers app
[21:06:31] <Mirja> mic: actually the AQM should not think about what higher layers are doing at all
[21:07:41] <wesley.m.eddy> mirja: we're going to try to move onto the presentation on evaluation (already behind schedule), but section 4.6 is going to be taken to the list, so you can weigh in there?
[21:07:54] <Mirja> yes
[21:08:04] <wesley.m.eddy> 6 is definitely still under discussion, following the meeting
[21:08:10] <Mirja> i think david said something similar
[21:13:36] rscheff joins the room
[21:19:15] <rscheff> slide 11
[21:21:27] <g.white> we experimented with it (SFQ-PIE) and can share an ns2 imll.
[21:21:47] <g.white> imll should be “implementation”
[21:21:54] <wesley.m.eddy> got it :)
[21:24:08] <rscheff> slide 15
[21:24:21] <Mirja> p_ecn < p_drop only work if you also change the congestion control as dctcp
[21:24:56] <Gabriel Zigelboim> @Mirja: < > ?
[21:25:26] <wesley.m.eddy> mirja: there's a mic queue 5-deep, and potentially someone in it will say this
[21:27:18] <g.white> in my view ecn / no-ecn coexistence and differential treatment in an AQM is NOT a solved problem (lot of open questions still)
[21:27:44] <Mirja> yes sorry >
[21:27:58] <Mirja> @wes:okay
[21:28:13] <Mirja> @greg: agree
[21:30:41] <wesley.m.eddy> not trying to filter you, but just want the people in the room to have time to blurt their comments out, before we loose the room
[21:31:09] <Mirja> i understand, that's fine
[21:39:44] rscheff leaves the room
[21:39:51] wesley.m.eddy leaves the room
[21:39:51] Gabriel Zigelboim leaves the room
[21:39:57] Mirja leaves the room
[21:40:55] jgunn leaves the room