Thursday, July 28, 2022< ^ >
jishac has set the subject to: Internet Congestion Control Research Group
Room Configuration
Room Occupants

[19:51:35] <zulipbot> (Gorry Fairhurst) It must be a great Beverage and Snack Break, there's no-one here !!! - I wonder what cookies were on offer?😀
[20:02:55] <zulipbot> (Gorry Fairhurst) Greetings Jana!
[20:04:17] <zulipbot> (Nick Banks) Yes
[20:04:30] <zulipbot> (Stephen Strowes) yes!
[20:04:30] <zulipbot> (Jonathan Lennox) I see your screen in the room
[20:04:56] <zulipbot> (Nick Banks) Yes, we can hear
[20:05:09] <zulipbot> (Asad Saeed) we can hear you
[20:07:33] <zulipbot> (Neal Cardwell) I'll take notes...
[20:07:46] <zulipbot> (Neal Cardwell) Yep.
[20:07:59] <zulipbot> (Neal Cardwell) Sure, you bet!
[20:09:24] <zulipbot> (Neal Cardwell) I'm not seeing Bob's slides...
[20:09:37] <zulipbot> (Neal Cardwell) Anyone else have this issue?
[20:09:37] <zulipbot> (Lucas Pardue) I had to reload the tab
[20:09:50] <zulipbot> (Richard Patterson) yeah, no slides here either
[20:10:03] <zulipbot> (Jonathan Morton) make sure you're in Presentation mode, leftmost of the top-right icons
[20:10:16] <zulipbot> (Jana Iyengar) The slides are online, you can follow along in case you can't get this working
[20:10:16] <zulipbot> (Jana Iyengar) We're on slide 3
[20:11:47] <zulipbot> (Neal Cardwell) Yep, I'm in presentation mode and just following along in the PDF
[20:12:00] <zulipbot> (Jana Iyengar) slide 4
[20:12:13] <zulipbot> (Hang Shi) What is ECN-Cubic? Does Cubic react to ECN?
[20:12:26] <zulipbot> (Neal Cardwell) Yes, it can. In Linux TCP it will.
[20:12:39] <zulipbot> (Jonathan Morton) it's just standard Linux CUBIC with ECN enabled, I assume
[20:13:05] <zulipbot> (Neal Cardwell) Yes, Linux CUBIC with RFC3168  ECN causing the cwnd reductions.
[20:13:18] <zulipbot> (Jonathan Morton) it's one of those things which "just works" if you flip the switch
[20:13:31] <zulipbot> (Hang Shi) Got it. Thx
[20:16:48] <zulipbot> (Hang Shi) Brilliant idea!
[20:19:01] <zulipbot> (Jonathan Morton) so far we've only been shown the ideal conditions; believe me things get more complicated than this
[20:23:19] <zulipbot> (Hang Shi) Will iOS 16 ship with TCP Prague?
[20:24:02] <zulipbot> (Jonathan Morton) I doubt it
[20:24:15] <zulipbot> (Matt Joras) fwiw this testing is with QUIC, not TCP afaik. I don't know if they actually implemented Prague in TCP?
[20:24:28] <zulipbot> (Jana Iyengar) Oh hm. I assumed this was TCP
[20:25:13] <zulipbot> (Jana Iyengar) Yup, QUIC.
[20:30:17] <zulipbot> (Neal Cardwell) what slide are we on now with Bob?
[20:30:18] <zulipbot> (Lucas Pardue) 8
[20:30:30] <zulipbot> (Neal Cardwell) thanks!
[20:32:47] <zulipbot> (Jonathan Morton) this is a really badly labelled set of graphs
[20:33:15] <zulipbot> (Neal Cardwell) Agreed ... :-)
[20:33:41] <zulipbot> (Jonathan Morton) slide 9
[20:36:13] <zulipbot> (Jana Iyengar) slide 10
[20:38:00] frodek joins the room
[20:38:05] <zulipbot> (Neal Cardwell) what's the metric graphed on the top row, again?
[20:38:18] <zulipbot> (Neal Cardwell) in slide 10?
[20:38:31] <zulipbot> (Jonathan Morton) in my experience, most of the "long tail" in the conventional TCPs' CDF plots is due to the startup transient
[20:39:10] <zulipbot> (Neal Cardwell) It would be useful to see the throughputs of the long-running flows in these cases?
[20:39:37] <zulipbot> (Jonathan Morton) I think that's the black bars at the bottom
[20:39:50] <zulipbot> (Jana Iyengar) yup
[20:40:03] <zulipbot> (Jonathan Morton) again, badly labelled graphs
[20:40:03] <zulipbot> (Neal Cardwell) but that only shows total utilization
[20:40:16] <zulipbot> (Jana Iyengar) yeah, it doesn't show distribution
[20:40:33] <zulipbot> (Jana Iyengar) Bob can respond when he returns to his laptop
[20:40:46] <zulipbot> (Jana Iyengar) It would be good to show that
[20:40:59] <zulipbot> (Neal Cardwell) Sounds great, thanks.
[20:41:12] <zulipbot> (Jonathan Morton) the results have been "stable since 2019" because no substantive technical changes have been made since then, despite a lot of pushback in TSVWG
[20:47:23] <zulipbot> (Jonathan Morton) my second question would have been to outline the requirements for the network path that can support a Prague CC
[20:47:36] <zulipbot> (Jonathan Morton) since there was absolutely no mention of that in the presentation
[20:47:49] <zulipbot> (Jonathan Morton) and it has been the single biggest issue in TSVWG
[20:51:11] <zulipbot> (Jonathan Morton) @Nathalie: if you move your microphone down a little, it should pop less
[20:51:33] <zulipbot> (Jonathan Morton) it is better now
[21:02:53] frodek leaves the room
[21:03:27] <zulipbot> (Spencer Dawkins) The question on Nathalie's last slide is perfect ...
[21:04:21] <zulipbot> (Markus Amend) +1
[21:06:09] frodek joins the room
[21:06:36] frodek leaves the room
[21:07:33] <zulipbot> (Neal Cardwell) Thanks, Jana!
[21:07:46] <zulipbot> (Jonathan Morton) Agreed with Colin's perspective here
[21:07:59] <zulipbot> (Neal Cardwell) Notes are at:
Powered by ejabberd - robust, scalable and extensible XMPP server Powered by Erlang Valid XHTML 1.0 Transitional Valid CSS!