[00:10:16] Christian Huitema joins the room
[03:24:28] Christian Huitema leaves the room: Disconnected: closed
[03:38:37] Christian Huitema joins the room
[03:41:14] Christian Huitema leaves the room: Disconnected: Replaced by new connection
[03:41:16] Christian Huitema joins the room
[04:20:23] Christian Huitema leaves the room: Disconnected: closed
[04:20:59] Christian Huitema joins the room
[04:21:39] Christian Huitema leaves the room: Disconnected: closed
[04:21:58] Christian Huitema joins the room
[10:20:42] alex-meetecho joins the room
[10:34:08] Meetecho joins the room
[10:47:01] lpardue joins the room
[10:50:01] Tobia Castaldi joins the room
[10:50:01] Alessandro Ghedini joins the room
[10:50:01] Marcus Ihlar joins the room
[10:50:01] Lucas Pardue joins the room
[10:50:01] Jiao Kang joins the room
[10:50:01] Marten Seemann joins the room
[10:50:01] Christopher Wood joins the room
[10:50:01] Zhuangyan joins the room
[10:50:35] Ted Hardie joins the room
[10:50:39] Alister Winfield joins the room
[10:52:23] Alister Winfield_479 joins the room
[10:52:40] Lucas Pardue leaves the room
[10:52:53] Magnus Westerlund joins the room
[10:52:55] Alex Chernyakhovsky joins the room
[10:53:05] <Tobia Castaldi> Hi everyone
[10:53:06] Lucas Pardue joins the room
[10:53:07] Christian Huitema leaves the room
[10:53:32] <Tobia Castaldi> If you have to present please make sure you follow the instructions described here: https://www.ietf.org/media/documents/Documentation-Meetecho-IETF.pdf
[10:53:41] Chi-Jiun Su joins the room
[10:53:46] Haizhou Du joins the room
[10:53:51] Dallas McCall joins the room
[10:53:59] Cullen Jennings joins the room
[10:54:08] Hirochika Asai joins the room
[10:54:12] Haizhou Du leaves the room
[10:54:37] Karl Kathuria joins the room
[10:55:05] Haizhou Du joins the room
[10:55:18] Haizhou Du leaves the room
[10:55:18] Haizhou Du joins the room
[10:55:19] Maksim Proshin joins the room
[10:55:36] Markus Amend joins the room
[10:55:56] Karl Kathuria_268 joins the room
[10:56:05] Jari Arkko joins the room
[10:56:38] Cullen Jennings_256 joins the room
[10:56:43] Tommy Pauly joins the room
[10:56:55] Spencer Dawkins joins the room
[10:56:58] Sean Turner joins the room
[10:57:11] Mirja Kühlewind joins the room
[10:57:53] Eckard Bogenfeld joins the room
[10:58:17] Akbar Rahman joins the room
[10:58:31] Alan Frindell joins the room
[10:59:02] Carrick joins the room
[10:59:06] spencerdawkins joins the room
[10:59:11] Nick Banks joins the room
[10:59:20] Jingwang Ma joins the room
[10:59:24] Alissa Cooper joins the room
[10:59:32] Alexandre Gouaillard joins the room
[10:59:34] Eric Kinnear joins the room
[10:59:35] Spencer Dawkins leaves the room
[10:59:35] Xavier de Foy joins the room
[10:59:35] Magnus Westerlund too joins the room
[10:59:36] Spencer Dawkins joins the room
[10:59:40] Philip Eardley joins the room
[10:59:43] Dan Druta joins the room
[10:59:44] Gorry Fairhurst joins the room
[10:59:45] Jordi Paillissé Vilanova joins the room
[10:59:46] Chris Box joins the room
[11:00:05] Marten Seemann leaves the room
[11:00:05] Marten Seemann joins the room
[11:00:05] Stephan Wenger joins the room
[11:00:12] Eric Kinnear leaves the room
[11:00:14] Nick Harper joins the room
[11:00:15] Benjamin Schwartz joins the room
[11:00:17] Eric Kinnear joins the room
[11:00:18] Jari Arkko leaves the room
[11:00:22] Alister Winfield_479 leaves the room
[11:00:25] ekinnear joins the room
[11:00:27] Murray Kucherawy joins the room
[11:00:29] Alister Winfield_917 joins the room
[11:00:32] Matthew Miller joins the room
[11:00:48] Maxime Piraux joins the room
[11:00:49] Renjie Tang joins the room
[11:00:58] Eric Rescorla joins the room
[11:01:02] Maxime Piraux leaves the room
[11:01:07] Maxime Piraux joins the room
[11:01:09] Philipp Tiesel joins the room
[11:01:10] John Border joins the room
[11:01:21] Greg White joins the room
[11:01:26] Martin Thomson joins the room
[11:01:46] Jari Arkko joins the room
[11:01:49] Kazuho Oku joins the room
[11:01:54] David Oliver joins the room
[11:02:01] Chris Wendt joins the room
[11:02:07] Mike Bishop joins the room
[11:02:20] David Schinazi joins the room
[11:02:28] Kirsty P joins the room
[11:02:31] Brian Trammell joins the room
[11:02:33] Martin Duke joins the room
[11:02:39] Jonathan Lennox joins the room
[11:02:43] Tom Jones joins the room
[11:02:47] <Martin Thomson> Oh great, video broken again.  This is great stuff.
[11:02:50] François Michel joins the room
[11:02:54] Craig Taylor joins the room
[11:02:59] Daniel Havey joins the room
[11:03:00] Chris Wendt leaves the room
[11:03:00] Chris Wendt joins the room
[11:03:20] Nicolas Kuhn joins the room
[11:03:22] Christopher Inacio joins the room
[11:03:25] Jasdip Singh joins the room
[11:03:33] Shogo Asaba joins the room
[11:03:44] Kiran Makhijani joins the room
[11:03:58] Jingwang Ma leaves the room
[11:04:11] dschinazi@jab.im joins the room
[11:04:15] Nicolas Kuhn leaves the room
[11:04:16] Jiao Kang leaves the room
[11:04:26] Brian joins the room
[11:04:29] Shogo Asaba leaves the room
[11:04:43] ihlar joins the room
[11:04:50] Jiao Kang joins the room
[11:05:03] Jingwang Ma joins the room
[11:05:08] <Mirja Kühlewind> Martin maybe you can give the meetecho team you IP address, so they can check your case specifically...?
[11:05:21] Alissa Cooper leaves the room
[11:05:21] Alissa Cooper joins the room
[11:05:58] Maxime Piraux leaves the room
[11:06:01] Nicolas Kuhn joins the room
[11:06:02] James Gould joins the room
[11:06:03] Martin Thomson leaves the room
[11:06:13] Martin Thomson joins the room
[11:06:18] Maxime Piraux joins the room
[11:06:19] <Tommy Pauly> Yeah, but he's hiding his IP address by sending his connections through a MASQUE proxy...
[11:06:23] achernya joins the room
[11:06:29] Chris Inacio joins the room
[11:06:32] <alex-meetecho> Martin Thomson: details that go beyond "video is broken" won't be completely useless
[11:06:35] <Benjamin Schwartz> I can take some notes
[11:06:52] <Ted Hardie> If you need something relayed, please put MIC: here in the jabber room.
[11:07:10] Marten Seemann leaves the room
[11:07:10] Marten Seemann joins the room
[11:07:22] Nicolas Kuhn leaves the room
[11:07:24] Martin Duke leaves the room
[11:07:30] Nicolas Kuhn joins the room
[11:07:34] <Martin Thomson> alex-meetecho: Where should I send a screenshot?
[11:07:54] Erik Nygren joins the room
[11:08:21] Michael Scharf joins the room
[11:08:29] Jari Arkko leaves the room
[11:08:47] Simon Romano joins the room
[11:08:59] <Martin Thomson> alex-meetecho: the recvonly video flow is connected and receiving bytes, but nothing displays
[11:09:00] <Meetecho> Martin Thomson: you can send it to ietf@meetecho.com thx, even though screenshots would only partly help. We'd need some more details like browser, possibly webrtc stats if you can gather them, etc.
[11:09:12] Bernard Aboba joins the room
[11:09:14] nygren joins the room
[11:09:30] <Meetecho> Martin Thomson: have you tried hovering on that video and hitting the pause button to stop it, and the play button to open it again?
[11:09:41] Craig Taylor leaves the room
[11:09:42] Sean Turner leaves the room
[11:09:46] Craig Taylor joins the room
[11:09:48] <Meetecho> What's the bitrate you see for that video?
[11:09:53] <Brian> i'm looking forward to fully-fledged protocol features that are accountable to lack of caffeine, actually...
[11:10:00] Sean Turner joins the room
[11:10:04] <Meetecho> If it's much less than, let's say, 200/300kbps it may be an issue another IETFer experienced
[11:10:40] Mo Zanaty joins the room
[11:11:09] Gorry Fairhurst leaves the room
[11:11:26] Martin Duke joins the room
[11:11:27] Gorry Fairhurst joins the room
[11:11:45] <Martin Thomson> 600kbps seems fine
[11:11:52] Simon Pietro Romano joins the room
[11:12:10] <Martin Thomson> no connectivity issues of note: srvflx candidates connected happily
[11:12:14] Jingwang Ma leaves the room
[11:12:32] Jingwang Ma joins the room
[11:12:34] Luca Niccolini joins the room
[11:12:53] <Meetecho> Ok so the browser is receiving the data but is either unable to decode it (missing keyframes?) or is not rendering it: have you checked if the HTML5 video element is playing? (right-clicking + show controls should confirm it)
[11:13:23] Jiao Kang leaves the room
[11:14:47] Jiao Kang joins the room
[11:14:56] Haizhou Du leaves the room
[11:14:58] Joerg Ott joins the room
[11:15:01] Haizhou Du joins the room
[11:15:14] <Martin Thomson> It showed the "Play" option, which I clicked to no effect.
[11:15:42] Dragana Damjanovic joins the room
[11:16:03] Florin joins the room
[11:16:11] <Meetecho> The pause/play trick in the controls in the upper/left part of the video didn't help either? Otherwise a page refresh may fix it
[11:16:25] <Mike Bishop> FWIW, on my (working) session, the same thing gives me a pause button.
[11:16:51] David Oliver leaves the room
[11:16:54] David Oliver joins the room
[11:17:16] Hannu Flinck joins the room
[11:17:29] Jiao Kang leaves the room
[11:17:30] Jiao Kang joins the room
[11:18:10] <Martin Thomson> I had the same problem yesterday, and the first session eventually started working with no intervention from me.
[11:18:20] <Martin Thomson> but I went through an entire session later with no video
[11:19:01] Craig Taylor leaves the room
[11:19:06] Craig Taylor joins the room
[11:19:30] Hirochika Asai leaves the room
[11:19:33] Hirochika Asai joins the room
[11:20:17] David Oliver leaves the room
[11:20:18] David Oliver joins the room
[11:20:23] Chathura Sarathchandra joins the room
[11:21:20] <spencerdawkins> We really need to pick a place to talk about nested congestion control - I'm guessing ICCRG? But that is a topic that's come up in multiple places.
[11:21:29] Anna Brunstrom joins the room
[11:21:39] <Martin Duke> "Considerations" for that are in the charter'
[11:21:46] Victor Vasiliev joins the room
[11:21:50] Gorry Fairhurst leaves the room
[11:21:51] Craig Taylor leaves the room
[11:21:52] Gorry Fairhurst joins the room
[11:21:57] Craig Taylor joins the room
[11:22:08] <Ted Hardie> If you want something relayed to the queue, pleae put MIC:  in the line you'd like relayed.
[11:22:28] <Martin Thomson> Not sure that I agree with Mike here, because streams provide reliability.
[11:22:32] Michael Scharf leaves the room
[11:22:33] Michael Scharf joins the room
[11:22:43] Craig Taylor leaves the room
[11:22:53] Craig Taylor joins the room
[11:22:54] David Oliver leaves the room
[11:23:04] David Oliver joins the room
[11:23:06] <Alan Frindell> not ready for unreliable qpack?
[11:23:24] <Martin Thomson> hah
[11:23:30] <Ted Hardie> @Alan We can call it unpack!
[11:23:30] David Oliver leaves the room
[11:23:34] David Oliver joins the room
[11:23:50] Craig Taylor leaves the room
[11:23:58] Craig Taylor joins the room
[11:24:18] <spencerdawkins> @Martin Duke - is nested congestion control close enough to engineering for the IETF? :-)
I was thinking that it still involved research, fairly recently.
[11:24:34] Tao Lin joins the room
[11:24:59] <Martin Duke> New algorithms are out of scope, but "recommendations" are in scope. The WG should write down something about this to avoid dumb stuff
[11:25:19] <Ted Hardie> I think the friction for supportng new connect methods is actually pretty high, even if we have a few at the moment.
[11:25:26] <Ted Hardie> So Methods make sense to me.
[11:25:51] <spencerdawkins> @Martin Duke - I'll probably say that in LOOPS on Friday :D
[11:26:11] Bhavit Shah joins the room
[11:26:20] <Martin Thomson> wow, video started.  All it took was for me to look at another tab for long enough, it seems
[11:27:06] Martin Duke leaves the room
[11:27:08] <Mirja Kühlewind> @Spencer: there is also https://datatracker.ietf.org/doc/draft-pauly-tsvwg-tcp-encapsulation/
[11:27:17] Bhavit Shah leaves the room
[11:27:24] Bhavit Shah joins the room
[11:27:34] <Daniel Havey> The Meetecho feature that shows the name is working now. But please say your name anyways.
[11:28:15] <ekinnear> We got a good bit of feedback and discussion on that draft in tsvwg, so not saying that’s the home for it, but at least we found _some_ of the right people :)
[11:28:31] <Mike Bishop> @MT, that's true, and would require retransmission logic for H3 frames.  But there are several frame types where you don't need to care about ordering.  I'm not necessarily saying we do it, but we've discussed it, so there may be value in not closing the door on them.
[11:28:31] Bhavit Shah leaves the room
[11:29:18] Tao Lin leaves the room
[11:29:19] Tao Lin joins the room
[11:29:21] <spencerdawkins> @Mirja - right. Tommy presented that in TSVAREA a while back, but the conversation then was about IKEv2 over TCP, and it wasn't obvious to me that it generalized. Thanks for the clue.
[11:29:56] Dan Druta leaves the room
[11:29:56] Dan Druta joins the room
[11:29:58] Martin Duke joins the room
[11:29:59] David Smith joins the room
[11:30:07] Jean-Michel Combes joins the room
[11:30:28] Jean-Michel Combes leaves the room
[11:31:18] Bhavit Shah joins the room
[11:32:07] Ian Swett joins the room
[11:32:10] Bhavit Shah leaves the room
[11:32:54] <Martin Thomson> So WebTransport seems like a strict superset of this in terms of functionality.
[11:33:42] <Martin Thomson> Mike Bishop: we talked about having stream-per-message for some of this so that we could get the reliability.  That was considered distasteful.
[11:33:43] <achernya> "this" being masque, or connect-udp, or the pattern slide?
[11:33:45] <Mirja Kühlewind> +1 to Magnus, I think it would be nice to work on UDP proxying and IP proxying and see how similar they are
[11:34:02] Joseph Salowey joins the room
[11:34:45] Jingwang Ma leaves the room
[11:34:51] Jingwang Ma joins the room
[11:34:57] <Victor Vasiliev> FWIW, the conceptual way I think about WebTransport over HTTP is "HTTP resource with TE that makes payload look like a QUIC connection", so it makes sense for it to feel like a superset
[11:35:11] <nygren > Will put here
[11:35:22] <spencerdawkins> New rule. At least one chair must be in a timezone where they're likely to be awake at any IETF virtual meeting ...
[11:35:35] dragana joins the room
[11:35:37] <Christopher Wood> Blame Meetecho.
[11:35:48] <ekinnear> Heh I like this plan
[11:35:54] Craig Taylor leaves the room
[11:35:58] <lpardue> what if I'm a nightowl?
[11:36:00] Craig Taylor joins the room
[11:36:05] <Meetecho> Christopher Wood: ?
[11:36:08] Stephan Emile joins the room
[11:36:14] <Martin Thomson> I like this, but I'm not sure about the proxy handling either.  David will recall a few issues being raised.
[11:36:16] <Ted Hardie> @spencer Since we can't travel, that could mean we need more chairs.
[11:36:16] <Mirja Kühlewind> I also had a audio error in another session. looks that this as well.
[11:36:45] <lpardue> +1 EKR
[11:36:49] <nygren > Another reason a CONNECTX might make sense would be for cross-protocol interactions.  In-particular, for handling ICMP(v6) responses to UDP messages such as for PMTUD/PTB.  Although the proxy may/will need to translate those regardless so it may be that that's a special-case and just having a general mechanism for relaying those back across the various CONNECT-* methods would also be fine.
[11:36:49] <Alan Frindell> +1
[11:36:56] <Mirja Kühlewind> meetecho Nygren was trying to join the queue but then got kick of audio immediately
[11:36:56] Bhavit Shah joins the room
[11:37:07] <Martin Thomson> The problem is that one proxy will pass an unknown method...
[11:37:11] Bhavit Shah leaves the room
[11:37:14] Jiao Kang leaves the room
[11:37:17] <Ted Hardie> @nygren I've enqueued to relay that.
[11:37:22] <spencerdawkins> @lpardue - all you have to do is be awake. @Ted, Lucas may have another solution :D
[11:37:31] Jiao Kang joins the room
[11:37:37] <Martin Thomson> Connection doesn't work in h2
[11:37:38] Craig Taylor leaves the room
[11:37:41] <Meetecho> Mirja Kühlewind: that can happen if permission to access the device was rejected by mistake (which Chrome remembers, for instance). In that case, permissions should be reset in the settings
[11:37:43] Craig Taylor joins the room
[11:37:53] <Eric Rescorla> I'm not there, but I would need to think about it in any case
[11:38:08] <Magnus Westerlund too> Chairs: Time check
[11:38:15] <Alan Frindell> Doesn't it have to be hop-by-hop?  Otherwise you could end up with collisions?
[11:38:19] <Eric Rescorla> Is Brian in disguise?
[11:38:23] <ekinnear> @Magnus ack
[11:38:43] <Mirja Kühlewind> meetecho I believe I had that problem when I only allowed for the mic but not the camera...
[11:39:05] <Martin Thomson> It's odd to see Brian during the daytime.
[11:39:18] <Christopher Wood> Cutting the queue here.
[11:39:24] <Mirja Kühlewind> @ekr super hot here today. I didn't even dare to sit outside...
[11:39:30] Chris Box leaves the room
[11:39:33] Chris Box joins the room
[11:39:53] <lpardue> it might be hot in Zurich but Brian looks cool
[11:40:25] Craig Taylor leaves the room
[11:40:30] Craig Taylor joins the room
[11:40:46] Erik Nygren leaves the room
[11:40:46] Erik Nygren joins the room
[11:41:06] <Mike Bishop> I think it's workable with or without an explicit proxy list -- the client can ask the (first) proxy to connect to the second proxy, or can request the end point and the proxy knows that's how it gets a UDP connection.
[11:42:00] chi.jiun.su joins the room
[11:42:18] <spencerdawkins> @Mike - +1
[11:42:30] <nygren > I think @meetecho is right — it looks like the browser I was using (Firefox) didn't have mic/video permissions in Mac OS X but wasn't usefully telling me this.
[11:42:57] Craig Taylor leaves the room
[11:43:03] Craig Taylor joins the room
[11:43:29] <Martin Thomson> So you can use this without an on-connection indication of support, but you degrade to a single-stream/single-connection in-ordered delivery design.  That's not bad, but you miss the upgrade to DATAGRAM.
[11:43:43] <spencerdawkins> +1 EKR - we've gotta stop calling everything a proxy no matter what it does. We talked about this at charter time.
[11:43:46] Craig Taylor leaves the room
[11:43:46] Craig Taylor joins the room
[11:43:52] <Meetecho> On MacOS you may have to also allow the browser to access those devices in the first place, for browsers that aren't safari: it's needed for screensharing in Catalina, for instance, but some other people needed to do the same for mic and webcam too (otherwise they're automatically rejected). The setting is usually found in the Privacy section of the OS
[11:44:33] <achernya> I suppose we're also confusing things by using a CONNECT-style method which evokes forward proxy but the diagram is showing a reverse roxy
[11:44:49] <Tommy Pauly> MASQUE should be forward proxies in general
[11:45:02] <Tommy Pauly> I think the diagram is confusing things =)
[11:45:06] <spencerdawkins> "protocol design is hard".
[11:45:06] <Martin Thomson> A setting to enable CONNECT-UDP seems like it is a reasonable strategy.  Then there is no confusion about what is going on.  It's very tricky if you allow DATAGRAM and CONNECT-UDP to be orthogonal negotiations.
[11:45:11] Maksim Proshin leaves the room
[11:45:11] <achernya> MASQUE is, but the implementation it runs over may involve reverse proxies :)
[11:45:21] Tao Lin leaves the room
[11:45:25] Tao Lin joins the room
[11:45:39] <Tom Jones> to left tab will a little plot
[11:45:41] <Mirja Kühlewind> but there is also the TOR like use case where you maybe want to talk to multiple (forward) proxies
[11:45:43] <Tom Jones> top*
[11:45:52] <Martin Thomson> what was that for?
[11:46:16] Sean Turner leaves the room
[11:46:20] Sean Turner joins the room
[11:46:23] <Martin Thomson> Oh, I would have hummed for "adopt", but I thought that it was a mistake.
[11:46:38] <Martin Thomson> We need a better voting tool.
[11:46:44] <Mirja Kühlewind> maybe we should redo the first hum
[11:46:47] <Eric Rescorla> well, who cares
[11:46:50] Haizhou Du leaves the room
[11:46:50] Haizhou Du joins the room
[11:46:54] <Eric Rescorla> because we have a hum for adoption anyway
[11:46:59] <spencerdawkins> I could have also gone for "adopt the next revision". It's really close, if it's not ready.
[11:47:01] <Martin Thomson> Yeah, what I want to know is who hummed against and why.
[11:47:12] <Eric Rescorla> VOTING!
[11:47:26] Yanyuan QIN joins the room
[11:47:40] <Mirja Kühlewind> actually on the mailing list we also "vote" anyway...
[11:47:55] <Erik Nygren> I didn't hum against, but concern would be if we want to take a different approach to better align with webtrans.
[11:47:59] <Martin Thomson> Vote early, vote often.
[11:48:17] <Eric Rescorla> Is he meaning to not show video?
[11:48:18] <lpardue> if we don't adopt the document and have no other proposal, then MASQUE is kinda a damp squib right?
[11:48:51] <Christopher Wood> Yeah, the hum was just to get a sense of direction in the room. We'll confirm and discuss reasons for and against adoption on the list.
[11:49:04] <Christopher Wood> (Responding to Mirja)
[11:49:17] <Martin Thomson> Erik, I share those concerns, but I have some amount of confidence that we can hold this until we're sure about the interplay between the two
[11:49:24] Kirsty P leaves the room
[11:49:26] Simon Vera-Schockner joins the room
[11:49:30] <ekinnear> @Erik Ack, I suspect we’ll have frequent communication with webtrans (especially as David is one of the webtrans chairs)
[11:49:30] Kirsty P joins the room
[11:49:33] <Ian Swett> +1 to Erik and MT's comments
[11:50:34] <Ian Swett> I assume "Indistinguishability" doesn't include traffic analysis?
[11:50:38] <Martin Thomson> Is the idea that you might use EAP over QUIC?
[11:50:38] Richard Barnes joins the room
[11:50:42] <Christopher Wood> I assume that as well, Ian.
[11:51:04] <Eric Rescorla> yes
[11:51:06] <achernya> I'm not personally fond of EAP itself, but a pluggable mechanism for mutual or client auth should exist
[11:51:12] <Martin Thomson> Yeah, I don't think that indistinguishability is practically achievable.
[11:51:13] <dschinazi@jab.im> @MT yes I think a requirement is for this to be extensible enough to allow building EAP over it as an extension
[11:51:54] <dschinazi@jab.im> And it depends on the definition of Indistinguishability
[11:52:15] Sanjay Mishra joins the room
[11:54:15] <achernya> 100% indistinguishable is an open research question, as there's work in identifying traffic through e.g., just length analysis
[11:54:32] <achernya> but I think we can do pretty well nonetheless
[11:56:06] <Martin Thomson> The flow control thing is weird.  That is something QUIC already does.
[11:56:21] <achernya> Datagram frames (IIRC) are not flow controlled, but streams are
[11:56:33] <Martin Thomson> Yes, that is right.
[11:56:39] <Martin Thomson> Did you want flow control for DATAGRAM?
[11:56:42] <achernya> We saw performance problems with QBONE until we turned QUIC flow control off entirely
[11:57:02] <Victor Vasiliev> Do you mean congestion control?
[11:57:06] <Martin Thomson> Sure, but that requires no protocol changes, just send 2^62-1 as your MAX_STREAM_DATA/MAX_DATA.
[11:57:09] <achernya> VictorL No, I mean both
[11:57:41] <Martin Thomson> Disabling congestion control seems reasonable, assuming that you still employ it end-to-end.
[11:57:42] <spencerdawkins> @Martin Thompson - I think it's weird, too. It seems like we're coming up with a lot of ways that allow you to do the same thing. Searching for a superset?
[11:57:45] <achernya> Martin: I think it's still worth documenting the expectation to be able to do so
[11:58:14] Carrick leaves the room
[11:58:22] Carrick joins the room
[11:58:54] <Martin Thomson> achernya: it is usual practice to list that differently than your unsatisfied requirements
[11:59:12] <Martin Thomson> I doubt that you can disable congestion control if you have to use HTTP/2
[11:59:32] <achernya> Martin: Apologies, as the datagram stuff is still a draft we wanted to call it out as a requirement
[11:59:36] <Eric Rescorla> It seems like if this is out of scope than the routing thing ought to be as well
[11:59:40] ekinnear has set the subject to: MASQUE @ IETF 108
[12:00:05] <achernya> ekr: you can do NAT with routing as a primitive
[12:00:10] <spencerdawkins> #Martin Thomson - right - you can TRY to disable congestion control for HTTP/2, but that should be a NOOP.
[12:00:11] <achernya> you can't do the converse
[12:00:25] <Eric Rescorla> achernya: I'm talking about "Addressing Architecture"
[12:00:27] <Martin Thomson> Agree with Ekr here regarding routing.  If the proxy is acting as a router, then it needs to be a router.  If it is a remote IP endpoint, then it isn't a router.
[12:00:37] <Ted Hardie> So this is a "success" if it proxies the packets over to a network, which then null routes them?
[12:00:50] <Eric Rescorla> Ted Hardie: my ISP would tell you that
[12:00:57] <achernya> ekr: I don't understand your point. We're trying to say in scope: some way to convey IP addresses/routes. Out of scope: deciding what those IPs are
[12:01:10] <Eric Rescorla> achernya: perhaps this is not the best venue for this
[12:01:36] ekr@jabber.org joins the room
[12:01:46] <spencerdawkins> I would say "thank you for doing this work" at the mike if we had more time ...
[12:01:59] Hannu Flinck leaves the room
[12:02:10] Alan Frindell leaves the room
[12:02:10] <Erik Nygren> Consideration here might be how to handle IPv6+IPv4 dual-stack environments, as well as transitions between IPv4 endpoints on side and IPv6 endpoints on the other side.
[12:02:23] <lpardue> I think indistinguishability should be a non-requirement. If its good enough (ignoring traffic pattern analysis) to just "hide" things in QUIC packets then that is the default behaviour, and you'd have to work hard to make the packet contents distinguishable
[12:02:24] <Ted Hardie> @Erik did you want that relayed?
[12:02:41] <Erik Nygren> Probably no need to relay at the moment.
[12:02:45] Renjie Tang leaves the room
[12:02:48] <Martin Thomson> spencerdawkins: consider this a reminder that maybe the "thank you" was always unnecessary.  I know it's good to provide positive feedback, but a mailing list is a good way to do that.
[12:02:56] <ekr@jabber.org> I think I agree with Ben that it would be helpful to go back and ask what the requirements are for other IP tunneling thngs
[12:03:03] Renjie Tang joins the room
[12:03:14] <ekr@jabber.org> So perhaps we need not reinvent
[12:03:23] Haizhou Du leaves the room
[12:03:30] Haizhou Du joins the room
[12:04:03] Brian Trammell leaves the room
[12:04:05] Leif Johansson joins the room
[12:05:25] Jiao Kang leaves the room
[12:05:46] Jiao Kang joins the room
[12:05:49] Rick Taylor joins the room
[12:05:59] Brian Trammell joins the room
[12:06:00] Haizhou Du leaves the room
[12:06:07] Haizhou Du joins the room
[12:06:16] <achernya> lpardue: That's not sufficient, because e.g., an ALPN would leak that you're doing ip proxying
[12:06:58] <lpardue> the TLS handshake is not HTTP
[12:07:41] <dschinazi@jab.im> @Lucas the Indistinguishability requirement is met by using h3
[12:07:57] Leif Johansson leaves the room
[12:08:40] <ekr@jabber.org> Well, if we're talking Tor, there is a bunch of prior art that uses fancy crypto, that this does not look like
[12:08:48] Alan Frindell joins the room
[12:08:48] <Martin Thomson> what does this do for PMTUD?
[12:09:43] <Mirja Kühlewind> I think the whole idea is to hide IP address from the server without all the fancy crypto
[12:09:49] Chris Box leaves the room
[12:09:49] Chris Box joins the room
[12:09:50] <Martin Thomson> This is not sufficient, sadly.  You need NEW_CONNECTION_ID at the level of the tunnel.
[12:10:13] <Magnus Westerlund too> You end up with two different, one for the packets being encapsualted and one for packets just forwarded with IP + UDP address+port rewriting
[12:10:14] <Mirja Kühlewind> yes you need a tunnel as well but then you don't need to all payload in the tunnel
[12:10:21] Bhavit Shah joins the room
[12:10:38] <Mirja Kühlewind> to send all payloads in the tunnel I mean
[12:11:06] <ekr@jabber.org> I'm having a lot of trouble following this, but if you want to have Tor-like properties, you need nested encryption
[12:11:14] <Martin Thomson> There is no draft for this presentation, right?
[12:11:21] <Mirja Kühlewind> no draft yet
[12:11:23] <Martin Thomson> Because this design doesn't work.
[12:11:34] <ekr@jabber.org> (though Tor uses tricks to avoid packet expansion)
[12:11:39] <Mirja Kühlewind> only some of the tor properties, that's why it's only tor-like
[12:11:40] <Martin Thomson> It would have been really good to get a draft so that we can discuss the design requirements.
[12:11:58] <ekr@jabber.org> Well, I have no idea what "tor-like" means
[12:12:14] <ekr@jabber.org> The place to start is with what the security objectives are
[12:12:25] <Martin Thomson> I can wait
[12:13:11] <Mirja Kühlewind> the main requirement is that no proxy on the path can see both the source IP address and the destination IP address
[12:13:13] Jiao Kang leaves the room
[12:13:17] Jiao Kang joins the room
[12:13:27] <ekr@jabber.org> but if the connection IDs are carried in the clear, then it doesn't help
[12:15:04] <Martin Thomson> I don't see much point in continuing a discussion about something that we can't see or discuss before the meeting.
[12:18:08] <ekr@jabber.org> Yeah, I think this is sufficiently intricate that its really hard to make sense of this way
[12:18:53] <Martin Thomson> There is a reasonable design to achieve this outcome, but it requires that you
[12:19:04] <Martin Thomson> have a whole new negotiation protocol.
[12:19:14] <Mirja Kühlewind> why?
[12:19:35] <Martin Thomson> because otherwise you have linkability for packets across the proxy, or an unmanageable proxy
[12:19:50] <Martin Thomson> the proxy has to be able to select the connection IDs that are included in packets that are sent to it
[12:20:03] <Mirja Kühlewind> or actually not sure what you mean by "a whole new"
[12:20:16] <Mirja Kühlewind> are we design a negotiation protocol here?
[12:20:35] <Martin Thomson> 5-tuple routing works, but it doesn't scale very well
[12:20:58] <Martin Thomson> it's another layer of negotiation in addition to the handshake stuff described for CONNECT-UDP
[12:21:39] <Martin Thomson> CONNECT (TCP) is not low overhead
[12:23:03] Zhuangyan leaves the room
[12:24:26] Alister Winfield_917 leaves the room
[12:24:41] Carrick leaves the room
[12:25:01] Carrick joins the room
[12:25:06] <Mirja Kühlewind> Ben with SOCKS you can send arbitrary cleartext but with this you have to use QUIC as it depends on the connection ID which is encrypted
[12:25:19] <Mirja Kühlewind> I mean quic is encrypted
[12:25:24] <Mirja Kühlewind> not the connID
[12:26:07] <Benjamin Schwartz> (Focused on note-taking right now.)
[12:26:54] James Gould leaves the room
[12:27:09] David Smith leaves the room
[12:28:43] Nicolas Kuhn leaves the room
[12:29:46] <Martin Thomson> flow label is not addressed, I would assume
[12:30:31] <Martin Thomson> As in, we might fix it per-"masque flow", which might do
[12:30:32] Chris Wendt leaves the room
[12:30:41] Chris Wendt joins the room
[12:30:55] <Martin Thomson> but we need to say something or it could be linkable
[12:31:30] Chathura Sarathchandra leaves the room
[12:31:36] <Martin Thomson> ECN seems like a good extension
[12:31:44] Marcus Ihlar leaves the room
[12:31:47] Marcus Ihlar joins the room
[12:31:59] <Chris Box> Re the skipped DSCPs: if we consider that Meetecho may eventually be streamed through MASQUE proxies I can see it would be very useful to have some kind of DSCP support through this chain, to allow suitable prioritisation of time-sensitive data.
[12:32:24] Ali Hussain joins the room
[12:32:40] Marcus Ihlar leaves the room
[12:32:40] Marcus Ihlar joins the room
[12:32:44] <Martin Thomson> Chris: this is ripe for extensions; think of adding multiple datagram-flow-id values for different DSCP codepoints
[12:33:02] Marcus Ihlar leaves the room
[12:33:07] <Tom Jones> why stop at 64k?
[12:33:09] Marcus Ihlar joins the room
[12:33:10] ihlar joins the room
[12:33:15] Haizhou Du leaves the room
[12:33:18] Haizhou Du joins the room
[12:33:22] <achernya> Probably because that's the maximum TLS record size, iirc
[12:33:25] <Martin Thomson> I assume that this has to fit in a UDP datagram
[12:33:32] <ekr@jabber.org> Max TLS record size is 2^14-1
[12:33:37] <Martin Thomson> Max TLS record size is 2^14
[12:33:45] <Martin Thomson> haha
[12:33:48] <Tom Jones> tls record size would have been a great argument to get rid of jumbograms
[12:33:55] <achernya> a bit smaller than I remembered, thanks
[12:33:57] Chunshan Xiong joins the room
[12:34:03] <Nick Harper> TLS records aren't used in QUIC
[12:34:17] <ekr@jabber.org> well that too, but it won't stop MT and I from correcting people :)
[12:34:38] Haizhou Du leaves the room
[12:34:38] <Martin Thomson> again, extensions seem appropriate
[12:34:46] Haizhou Du joins the room
[12:34:59] Tao Lin leaves the room
[12:35:04] Haizhou Du leaves the room
[12:35:05] Tao Lin joins the room
[12:35:36] Martin Duke leaves the room
[12:35:55] Jiao Kang leaves the room
[12:35:58] Jingwang Ma leaves the room
[12:37:01] <Mirja Kühlewind> I would at least like to see native support for ECN :-)
[12:37:06] <ekr@jabber.org> Maybe we should have some extensions
[12:37:25] <dschinazi@jab.im> I think ECN should be the first extension
[12:37:32] <Ian Swett> You can proxy ECN back to the other endpoint, that seems sufficient for v1.
[12:38:02] <Benjamin Schwartz> Proxy ECN raises privacy concerns
[12:38:03] <Ian Swett> (nevermind, that's connect-IP, not connect-UDP)
[12:38:14] Chris Wendt leaves the room
[12:38:19] Chris Wendt joins the room
[12:38:22] <Benjamin Schwartz> Oh, ok
[12:38:23] Bhavit Shah leaves the room
[12:38:45] Tao Lin leaves the room
[12:38:48] Tao Lin joins the room
[12:39:03] Tao Lin leaves the room
[12:39:12] Tao Lin joins the room
[12:39:27] Tao Lin leaves the room
[12:39:37] Tao Lin joins the room
[12:39:40] Chris Wendt leaves the room
[12:39:52] Tao Lin leaves the room
[12:40:15] <Mirja Kühlewind> I actually just realised that we might have different few on if there is (another) IP header in the tunnel in the case of IP proxying
[12:40:22] <Jonathan Lennox> If you're doing ECN in a UDP application, you need to use the RECVTOS sockopt or the like.
[12:40:28] <achernya> IP, or UDP?
[12:40:42] <Mirja Kühlewind> for ups connect we only have quic over quic over udp over ip
[12:40:59] <Jonathan Lennox> UDP - the OS doesn't do any ECN support for UDP.
[12:41:06] <Mirja Kühlewind> so I would assume ip proxying would be .e.g quic over udp over quic over udo over ip
[12:41:11] <achernya> We've had discussion about CONNECT-UDP, requirements for CONNECT-IP, and a high-level design for CONNECT-QUIC
[12:41:19] <achernya> Only CONNECT-IP has inner IP headers here right now
[12:41:32] <Tom Jones> tos field support is patchy across operating systems
[12:41:49] <Tom Jones> linux and bsd support it, macos and windows don't (or didnt')
[12:41:50] <spencerdawkins>   If masque doesn't support a lot of this functionality as baseline, most of the extensions may not be useful in actual deployments ...
[12:41:52] <Mirja Kühlewind> I was assuming that you don't have an inner IP header but you are adding one before you send it to the target
[12:42:10] <Martin Thomson> spencerdawkins: let the market decide
[12:42:13] <achernya> That's true for CONNECT-UDP and CONNECT-QUIC, but not CONNECT-IP
[12:43:06] <Martin Thomson> set DF always as a requirement of the protocol
[12:43:09] <spencerdawkins> @Martin Thomson: - I believe you, but I expect the result to be mostly unusable in general deployments
[12:43:20] Ali Hussain leaves the room
[12:43:46] <achernya> +1 to tunneling the entire IP packet and not using CONNECT-UDP if you have these requirem,ents
[12:43:59] <Jonathan Lennox> macOS has RECVTOS
[12:44:04] Simon Romano leaves the room
[12:44:07] Nick Banks leaves the room
[12:44:09] Ian Swett leaves the room
[12:44:09] Philip Eardley leaves the room
[12:44:10] Ted Hardie leaves the room
[12:44:11] <Jonathan Lennox> I think everything modern but Windows has it
[12:44:11] Eric Rescorla leaves the room
[12:44:11] Victor Vasiliev leaves the room
[12:44:11] Jasdip Singh leaves the room
[12:44:12] Alissa Cooper leaves the room
[12:44:14] Luca Niccolini leaves the room
[12:44:15] Mike Bishop leaves the room
[12:44:15] Chi-Jiun Su leaves the room
[12:44:16] Nick Harper leaves the room
[12:44:16] Eckard Bogenfeld leaves the room
[12:44:16] Brian Trammell leaves the room
[12:44:17] Marcus Ihlar leaves the room
[12:44:17] Carrick leaves the room
[12:44:18] Renjie Tang leaves the room
[12:44:18] Tommy Pauly leaves the room
[12:44:18] Daniel Havey leaves the room
[12:44:19] Cullen Jennings_256 leaves the room
[12:44:19] Sanjay Mishra leaves the room
[12:44:19] Dallas McCall leaves the room
[12:44:19] Anna Brunstrom leaves the room
[12:44:21] David Oliver leaves the room
[12:44:21] Maxime Piraux leaves the room
[12:44:21] Matthew Miller leaves the room
[12:44:21] Murray Kucherawy leaves the room
[12:44:22] Greg White leaves the room
[12:44:23] François Michel leaves the room
[12:44:23] <Mirja Kühlewind> thanks! bye!
[12:44:24] Xavier de Foy leaves the room
[12:44:24] Alan Frindell leaves the room
[12:44:24] Joerg Ott leaves the room
[12:44:27] Alex Chernyakhovsky leaves the room
[12:44:28] Dragana Damjanovic leaves the room
[12:44:28] Eric Kinnear leaves the room
[12:44:28] Kirsty P leaves the room
[12:44:29] Karl Kathuria_268 leaves the room
[12:44:30] John Border leaves the room
[12:44:31] Sean Turner leaves the room
[12:44:31] Dan Druta leaves the room
[12:44:32] Joseph Salowey leaves the room
[12:44:36] Florin leaves the room
[12:44:37] Jordi Paillissé Vilanova leaves the room
[12:44:37] Chris Box leaves the room
[12:44:39] Jonathan Lennox leaves the room
[12:44:43] Gorry Fairhurst leaves the room
[12:44:43] Tobia Castaldi leaves the room
[12:44:43] Mo Zanaty leaves the room
[12:44:43] Richard Barnes leaves the room
[12:44:43] Martin Thomson leaves the room
[12:44:43] Magnus Westerlund leaves the room
[12:44:43] Hirochika Asai leaves the room
[12:44:43] Stephan Wenger leaves the room
[12:44:43] Spencer Dawkins leaves the room
[12:44:43] David Schinazi leaves the room
[12:44:43] Mirja Kühlewind leaves the room
[12:44:43] Marten Seemann leaves the room
[12:44:43] Yanyuan QIN leaves the room
[12:44:43] Benjamin Schwartz leaves the room
[12:44:43] Bernard Aboba leaves the room
[12:44:43] Craig Taylor leaves the room
[12:44:43] Akbar Rahman leaves the room
[12:44:43] Lucas Pardue leaves the room
[12:44:43] Erik Nygren leaves the room
[12:44:43] Simon Vera-Schockner leaves the room
[12:44:43] Tom Jones leaves the room
[12:44:43] Christopher Inacio leaves the room
[12:44:43] Philipp Tiesel leaves the room
[12:44:43] Kiran Makhijani leaves the room
[12:44:43] Christopher Wood leaves the room
[12:44:43] Rick Taylor leaves the room
[12:44:43] Markus Amend leaves the room
[12:44:43] Chunshan Xiong leaves the room
[12:44:43] Alessandro Ghedini leaves the room
[12:44:43] Alexandre Gouaillard leaves the room
[12:44:43] Michael Scharf leaves the room
[12:44:43] Stephan Emile leaves the room
[12:44:43] Kazuho Oku leaves the room
[12:44:46] spencerdawkins leaves the room
[12:44:59] alex-meetecho leaves the room
[12:46:28] dschinazi@jab.im leaves the room
[12:46:36] Magnus Westerlund too leaves the room
[12:47:21] achernya leaves the room
[12:47:30] ihlar leaves the room
[12:48:35] Simon Pietro Romano leaves the room
[12:49:42] Meetecho leaves the room
[12:49:44] ihlar leaves the room
[12:57:37] nygren leaves the room
[12:58:46] Brian leaves the room
[13:00:33] Chris Inacio leaves the room
[13:09:18] Alister Winfield leaves the room
[13:09:45] ekinnear leaves the room
[13:11:50] Karl Kathuria leaves the room: Disconnected: closed
[13:12:20] Simon Pietro Romano joins the room
[13:15:25] erik joins the room
[13:35:52] Christian Huitema joins the room
[13:44:22] nygren joins the room
[13:54:11] Cullen Jennings leaves the room
[14:25:53] lpardue leaves the room: Disconnected: BOSH client silent for over 60 seconds
[15:57:07] ekr@jabber.org leaves the room
[16:18:13] Christian Huitema leaves the room: Disconnected: closed
[17:41:20] nygren leaves the room
[17:57:29] dragana leaves the room: Disconnected: closed
[18:51:40] Simon Pietro Romano leaves the room
[18:51:57] Simon Pietro Romano joins the room
[19:22:40] Simon Pietro Romano leaves the room
[23:13:14] chi.jiun.su leaves the room