IETF
tsvwg
tsvwg@jabber.ietf.org
Friday, July 26, 2019< ^ >
craigt has set the subject to:
Room Configuration
Room Occupants

GMT+0
[16:09:31] Meetecho joins the room
[16:09:31] VirtualQueue_dt3oTbuP leaves the room
[16:15:06] Joakim Misund joins the room
[16:15:08] Wesley Eddy joins the room
[16:15:10] John Leslie joins the room
[16:16:31] Sebastian Moeller joins the room
[16:19:23] Michael Scharf joins the room
[16:20:08] Kevin Smith joins the room
[16:20:56] Jonathan Lennox joins the room
[16:21:22] Kevin Smith leaves the room
[16:21:39] Yingzhen Qu joins the room
[16:22:55] Olivier Tilmans joins the room
[16:23:18] Kevin Smith joins the room
[16:23:39] craigt joins the room
[16:25:29] Magnus Westerlund joins the room
[16:25:43] Neal Cardwell joins the room
[16:26:06] Sebastian Moeller leaves the room
[16:26:13] Sebastian Moeller joins the room
[16:26:43] Dave Täht joins the room
[16:27:02] jholland joins the room
[16:27:23] Sebastian Moeller leaves the room
[16:27:24] Sebastian Moeller joins the room
[16:28:23] Joakim Misund leaves the room
[16:28:24] Joakim Misund joins the room
[16:28:26] Sebastian Moeller leaves the room
[16:28:27] Dave Täht leaves the room
[16:28:28] Sebastian Moeller joins the room
[16:29:12] dave taht joins the room
[16:31:47] Sebastian Moeller leaves the room
[16:31:50] Sebastian Moeller joins the room
[16:33:42] Praveen Balasubramanian joins the room
[16:35:40] Nicholas Banks joins the room
[16:40:37] Ram Ranganathan joins the room
[16:42:50] loganaden velvindron joins the room
[16:43:33] <loganaden velvindron> hello
[16:44:06] <jholland> hi
[16:44:42] krose joins the room
[16:45:50] <dave taht> hi all
[16:46:32] <Sebastian Moeller> Hi
[16:47:33] Nicholas Banks leaves the room
[16:49:02] <loganaden velvindron> i would liike to know the CDNs
[16:49:13] <loganaden velvindron> if they are ok with identifying themselves
[16:49:19] <loganaden velvindron> and who to talk to
[16:49:21] <jholland> i have observed some ce marks on the internet.  i'll go to the mike.
[16:49:22] <dave taht> I figure CE is coming from fq_codel enabled containers in the cloud at this point.
[16:49:26] <jholland> i should say ece
[16:49:48] nygren joins the room
[16:52:00] Stephanie Huguenin joins the room
[16:52:53] <dave taht> inbound shaping is almost uiversally used in the sqm deployment.
[16:53:15] <jholland> yes, evenroute automates it.  it's pretty sweet, actually.
[16:53:31] <Sebastian Moeller> Unfortunatelly we have exactly zero numbers for deployed devices and/or activated ingress sqm instances
[16:53:49] <dave taht> My networks show (with the inbound shaping in place on comcast) - about 30x more CE marks than outbound.
[16:54:33] <dave taht> it's still at a low level (roughly 5% CE vs 95% drop)
[16:54:45] <dave taht> but up from zero 2 years back.
[16:54:59] <craigt> jabber: Please preface with "Mic:" anything you want me to relay
[16:55:20] <loganaden velvindron> thanks jake for answering this question.
[16:55:23] Tom Henderson joins the room
[16:55:52] <dave taht> haven't had anything so far, aside from having never been able to reproduce the dctcp results in the first slide, and I'd talked to reproducability problem so much I don't feel there's a need to go to the mic for it.
[16:56:49] <dave taht> the fact that "my" dctcp implementation had no response to loss at all was a revalation.
[16:58:19] <dave taht> My "CE" result is actually on the border gateway, it's highly variable and spread also through my wifi deployment.
[16:59:06] <dave taht> Ranging from 0 to nearly total, depending on how bad the link.
[16:59:33] <dave taht> still not questions for the mic, though. Getting organized about the wifi stuff will take more effort and $ than I have.
[17:00:03] Ram Ranganathan leaves the room
[17:00:55] Daniel Havey joins the room
[17:02:33] loganaden velvindron leaves the room
[17:03:36] loganaden velvindron joins the room
[17:05:53] <dave taht> https://arxiv.org/pdf/1903.03852.pdf - 50 shades of congestion control - is a good comparison for delay and other congestion controls
[17:05:54] Juan P. Cerezo joins the room
[17:06:24] <craigt> Jonson Watt?
[17:08:05] nygren leaves the room
[17:08:20] <Magnus Westerlund> Jonathan Morton at the mic currently
[17:08:28] <craigt> Thank you :)
[17:20:23] Kevin Smith leaves the room
[17:22:28] Stephanie Huguenin leaves the room
[17:24:33] Praveen Balasubramanian leaves the room
[17:25:29] <dave taht> "classic" vs "normal"
[17:26:09] <krose> It could be "legacy" ;-)    "Classic" isn't too bad.
[17:27:04] <dave taht> hahaha! yes, the hype rankles.
[17:28:40] <dave taht> read a recent version
[17:29:12] <dave taht> yes
[17:29:14] <dave taht> yes
[17:29:36] Alistair Woodman joins the room
[17:30:24] <Michael Scharf> Dave: Which ones?
[17:30:32] <Michael Scharf> (Asking as notetaker=
[17:30:46] <dave taht> Hmm? Read 'em all. There are 20 remote participants
[17:31:17] <dave taht> have no idea who attending has read.
[17:32:10] <jholland> there were ~15-20 in the room who raised hands for having read the docs
[17:32:15] <jholland> more for arch than for the other 2
[17:32:23] <dave taht> Ibut none in this room.
[17:32:42] <dave taht> humm
[17:32:49] <dave taht> (not wglc)
[17:32:56] <craigt> ack
[17:36:57] <Michael Scharf> @craigt: Is this a hum? If so, for what?
[17:37:19] Alistair Woodman leaves the room
[17:38:52] Alistair Woodman joins the room
[17:41:46] krose leaves the room
[17:45:51] <craigt> Sorry Michael Scharf: That was a hum for 1) wglc on L4S collection or 2) Not ready for WGLc
[17:46:33] <craigt> Outcome was unclear; Magnus asked for clear technical issues to the collection to be raised on the list
[17:47:29] <craigt> Over the next 'few weeks'. WGLC would proceed without such objections to proceeding being raised.
[17:47:39] <Magnus Westerlund> I wouldn't call the outcome as unclear. The hums where very similar in strength, so no even rough consensus. And thus lets try to address the issues why it isn't ready.
[17:47:51] <Wesley Eddy> FYI - I am planning to start an issue tracker, populated based on the discussion today
[17:48:17] <Magnus Westerlund> I think that is a very good idea Wes.
[17:48:17] <craigt> s/unclear/no clear consensus/
[17:49:15] <craigt> Wesley Eddy: That's works really well elsewhere
[17:59:35] craigt leaves the room
[17:59:48] craigt joins the room
[17:59:53] <dave taht> could you flip back to one of single queues graph
[18:00:46] <dave taht> thx!
[18:01:33] <dave taht> as for the low end increase in the delay, that is an artifact of a real linux system, as opposed to a simulation.
[18:02:08] <dave taht> A similar flent test will be similar for l4s-dctcp, but we don't have those.
[18:02:19] Neal Cardwell leaves the room
[18:02:25] craigt leaves the room: Disconnected: closed
[18:04:09] <dave taht> would kind of prefer to be back looking at the pretty graph. :)
[18:04:29] <dave taht> ut it is a good point to compare reno normal vs cubic normal for reference. flent can do that easily.
[18:06:34] Daniel Havey leaves the room
[18:06:43] <Sebastian Moeller> SCE seems to be fully compatible with L4S just not with the use of ECT(1) as classifying heuristic
[18:07:39] <dave taht> WOW! Can I get some of that hardware!!!!!?
[18:07:44] Michael Scharf leaves the room
[18:09:54] Olivier Tilmans leaves the room
[18:11:04] Magnus Westerlund leaves the room: Stream reset by peer
[18:11:22] <jholland> @sebastian: sort of. i think generally the objections to that idea are instead based on the deployability of the alternatives for classification (basically either fq or dscp, both of which have their own problems), but i do agree people mostly agree it's probably viable to the extent dscp is viable, and it would (to me) address the signaling concerns I've raised.
[18:12:07] John Leslie leaves the room
[18:12:10] Joakim Misund leaves the room
[18:13:15] Tom Henderson leaves the room
[18:17:57] <dave taht> I think l4s inevitably will end in ISPs bleaching out external ecn enabled traffic entirely (both codepoints) in favor of their own traffic. SCE (as it explicitly is asking for less priority) is less likely.
[18:18:17] <Meetecho> How long is this session going to last? We do need to wrap up our equipment soon...
[18:19:14] <Meetecho> If anyone is in the room, please ask the chairs, and notify them we might have to close the remote participation service before it's done
[18:19:28] <jholland> will do
[18:20:14] <Meetecho> (y)
[18:20:22] <Meetecho> Thanks for relaying!
[18:20:57] <jholland> np
[18:21:08] jholland leaves the room
[18:22:26] Yingzhen Qu leaves the room
[18:22:30] Wesley Eddy leaves the room
[18:22:34] dave taht leaves the room
[18:22:34] loganaden velvindron leaves the room
[18:22:34] Sebastian Moeller leaves the room
[18:22:34] Alistair Woodman leaves the room
[18:22:34] Juan P. Cerezo leaves the room
[18:22:37] Jonathan Lennox leaves the room
[18:22:37] Meetecho leaves the room
Powered by ejabberd - robust, scalable and extensible XMPP server Powered by Erlang Valid XHTML 1.0 Transitional Valid CSS!