IETF
QUIC
quic@jabber.ietf.org
Thursday, March 30, 2017< ^ >
Sean Turner has set the subject to: QUIC Interim - 20170124-20170126 | 9:30 am  |  Japan Time (Tokyo, GMT+09:00)  |  8 hrs
Room Configuration
Room Occupants

GMT+0
[00:29:50] ekaterina leaves the room
[00:54:18] ekaterina joins the room
[01:32:12] ekaterina leaves the room
[04:46:32] fujisaki.tomohiro joins the room
[04:46:40] fujisaki.tomohiro leaves the room
[11:17:19] ekaterina joins the room
[11:21:09] ekaterina leaves the room
[12:42:20] ekaterina joins the room
[12:43:14] ekaterina leaves the room
[13:11:18] Samuel Jero joins the room
[13:43:13] Meetecho joins the room
[13:43:37] ekaterina joins the room
[13:51:50] Yoshiro Yoneya joins the room
[13:55:08] Natasha Rooney joins the room
[13:55:08] Kaoru Maeda joins the room
[13:55:09] Lucas Clemente joins the room
[13:55:09] a a joins the room
[13:55:09] Olivier Bonaventure joins the room
[13:55:24] Marten Seemann joins the room
[13:56:12] Olivier Bonaventure leaves the room
[13:56:33] Edward Lopez joins the room
[13:56:52] Iheanyi Obasi joins the room
[13:57:05] Samuel Jero_ joins the room
[13:58:13] s d joins the room
[13:58:13] Colin Perkins joins the room
[13:58:19] Matthew Ford joins the room
[13:58:25] Martin Thomson joins the room
[13:58:26] Alain Durand joins the room
[13:58:48] Dale Worley joins the room
[13:58:57] John Leslie joins the room
[13:59:02] <Matthew Ford> I'm scribing - if you would like a comment or question relayed at the mic, please prefix your comment with MIC:
[13:59:05] Simon Pietro Romano joins the room
[13:59:15] Olivier Bonaventure joins the room
[13:59:30] Aaron Falk joins the room
[13:59:36] Simon Pietro Romano leaves the room
[13:59:48] Marten Seemann leaves the room
[13:59:52] Marten Seemann joins the room
[13:59:53] Simon Pietro Romano joins the room
[14:00:14] Ken Murchison joins the room
[14:00:15] Francis Teague joins the room
[14:00:38] craigt joins the room
[14:00:48] Charles Krasic joins the room
[14:01:11] p m joins the room
[14:01:14] Jim Welch joins the room
[14:01:21] Julian Reschke joins the room
[14:01:33] Spencer Dawkins joins the room
[14:01:45] Yang Song joins the room
[14:01:46] ajs joins the room
[14:02:10] Buck joins the room
[14:02:12] Chi-Jiun Su joins the room
[14:02:17] <Matthew Ford> We are starting - Mark Nottingham is reviewing the agenda
[14:02:41] yuki goto joins the room
[14:03:05] kadukoafs@gmail.com/barnowlD26A4D0C joins the room
[14:03:13] Alain Durand_3681 joins the room
[14:03:27] Martin Duke joins the room
[14:03:28] lucas_pardue joins the room
[14:03:38] g.e.montenegro joins the room
[14:03:56] Chi-Jiun Su_4101 joins the room
[14:03:58] dkg joins the room
[14:04:18] Daniel Stenberg joins the room
[14:04:18] <Matthew Ford> Lars Eggert introduces the meeting room setup experiment
[14:04:20] Chi-Jiun Su leaves the room
[14:04:28] Richard Bradbury joins the room
[14:04:47] <Matthew Ford> Mark presents a status report of the WG
[14:05:08] ted.h joins the room
[14:05:27] Dapeng Liu joins the room
[14:05:40] Ronald van der Pol joins the room
[14:05:42] Okke Timm joins the room
[14:05:42] <Matthew Ford> Now that we've started, I'll say this again
[14:05:44] <Matthew Ford> I'm scribing - if you would like a comment or question relayed at the mic, please prefix your comment with MIC:
[14:05:51] svaldez@tls.ctf.network joins the room
[14:06:02] Valery Smyslov joins the room
[14:06:23] Safa Almalki joins the room
[14:06:25] Fan Yang joins the room
[14:07:06] koji joins the room
[14:07:17] Padma Bhooma joins the room
[14:07:19] julien maisonneuve joins the room
[14:07:22] Bob Briscoe joins the room
[14:08:06] cameron byrne joins the room
[14:08:28] Wendy Seltzer joins the room
[14:09:37] Vicky Risk joins the room
[14:09:44] <Julian Reschke> who's at the mic?
[14:09:47] Alain Durand_7025 joins the room
[14:09:50] nygren joins the room
[14:10:25] <kadukoafs@gmail.com/barnowlD26A4D0C> (I'm sure Mark knows about the githbu permissions issues w.r.t. who is
allowed to apply tags.)
[14:10:50] Safa Almalki_3482 joins the room
[14:11:20] <Martin Thomson> kaduk, I'm sure that Mark is aware, we are OK with only editors being able to apply tags
[14:11:28] <Martin Thomson> in fact, that's a good state in my opinion
[14:12:13] <kadukoafs@gmail.com/barnowlD26A4D0C> Martin: right.  I started to get up to say that but realized it was
not a useful use of time.
[14:13:11] <Martin Thomson> if you feel inclined to say anything again, there is a seat next to DKG
[14:13:58] Arash Molavi Kakhki joins the room
[14:13:59] <kadukoafs@gmail.com/barnowlD26A4D0C> I mean, I was not planning to have much to say, so I didn't sit at the
table deliberately.  But duly noted.
[14:15:20] <Colin Perkins> Jana is very hard to understand on the remote audio
[14:15:24] <Buck> +1
[14:15:32] <Yang Song> +1
[14:15:34] <Samuel Jero_> Yeah. It sounds to me like we don't have that mic remotely
[14:16:04] <craigt> Meetecho: did you see the above?
[14:16:11] Safa Almalki_3482 leaves the room
[14:16:19] <Simon Pietro Romano> Is he talking into the mic?
[14:16:34] <kadukoafs@gmail.com/barnowlD26A4D0C> He's talking into a table mic, yes.
[14:16:40] <Martin Thomson> in the room, Jana is quite loud
[14:16:43] <craigt> Yes, he's mic'd in the room
[14:16:45] safa almalky joins the room
[14:17:09] <Simon Pietro Romano> OK. looking into that.
[14:17:14] <ted.h> I think that having it in a different document would be useful, because it may adjust as we add using protocols.
[14:17:33] Satoru Kanno joins the room
[14:17:33] <Buck> can Jana change seats?
[14:17:45] <ted.h> Our sense of the API may be very different once we have partial reliability
[14:17:49] <Natasha Rooney> mic
[14:17:59] safa almalky leaves the room
[14:18:58] Stanislav Slusny joins the room
[14:19:09] <Samuel Jero_> Not remotely
[14:19:11] <Buck> i can not hear Jana still
[14:19:19] <Samuel Jero_> That mic is not piped into the remote feed
[14:19:19] <Colin Perkins> Can't hear Jana
[14:19:23] Randell Jesup joins the room
[14:19:26] <Arash Molavi Kakhki> We can't hear it well remotely !
[14:19:31] <Samuel Jero_> We hear him echoing in other mics in the room
[14:19:45] <Simon Pietro Romano> Somebody from Meetecho is coming there
[14:19:55] <kadukoafs@gmail.com/barnowlD26A4D0C> Can y'all hear Sean's 'testing'?
[14:20:00] frodek joins the room
[14:20:07] <Buck> in the background
[14:20:10] SAfe malki joins the room
[14:20:11] <Colin Perkins> Was very quiet
[14:20:14] <Samuel Jero_> Same problem. His mic isn't piped into the remote feed
[14:20:16] <Jim Welch> no
[14:20:19] <Spencer Dawkins> Meetecho thank you.
[14:20:21] <Meetecho> Yep, working on that
[14:20:23] <Matthew Ford> so maybe all the table mics have this problem?
[14:20:27] <Simon Pietro Romano> Just 40 floors :-)
[14:20:29] <Matthew Ford> do you hear lars and mark clearly?
[14:20:31] <ted.h> Could you hear dkg earlier?
[14:20:39] <Samuel Jero_> Possibly? Yes, chairs were clear
[14:20:48] <Colin Perkins> I hear Lars, Mark, DKJ, Mirja, and Gorry well
[14:20:50] <Meetecho> It should only be one or a few mics, as we tested them before
[14:20:55] <Meetecho> sorry, new configuration for this room today
[14:21:00] <Matthew Ford> mirja and gorry where using floor mics
[14:21:39] Kyle Rose joins the room
[14:22:56] Julian Reschke leaves the room
[14:23:05] Julian Reschke joins the room
[14:24:42] <Meetecho> Let us know if audio has improved
[14:25:17] JoeHallCDT joins the room
[14:25:17] <Matthew Ford> Jana is now using floor mic, so should be audible
[14:25:22] <Colin Perkins> yes
[14:25:38] Magnus Westerlund joins the room
[14:25:38] <kadukoafs@gmail.com/barnowlD26A4D0C> He also moved to a different spot at the table, so will be on a
different table mic when he goes back to his seat.
[14:25:46] <Kyle Rose> At least the list isn't in pink
[14:25:56] <kadukoafs@gmail.com/barnowlD26A4D0C> We might be able to get Sean to do a mic check in between
presentations, though.
[14:25:58] <Martin Thomson> are WINDOW_UPDATE frames really fatter than the other packets?
[14:26:37] <Meetecho> there were 2 mics muted on the mixer, which should have been fixed now
[14:26:42] <Meetecho> let us know if that's not the case
[14:27:27] <kadukoafs@gmail.com/barnowlD26A4D0C> Ah.  We should probably not interrupt them right now just to check, so
we'll name you once we have better data either way.
[14:27:50] Dan York joins the room
[14:28:10] <Meetecho> sure, didn't mean to disrupt :) just wanted to update on the status
[14:28:19] <kadukoafs@gmail.com/barnowlD26A4D0C> Sure
[14:29:09] SAfe malki leaves the room
[14:29:53] sam noman joins the room
[14:29:58] <kadukoafs@gmail.com/barnowlD26A4D0C> nygren, you're sitting in front of one of the affected mics; will you
say something like "mic check for the remote attendees; tell the
jabber room if you can hear me" at the next break between
presentations?
[14:32:55] Martin Duke leaves the room
[14:33:01] Martin Duke joins the room
[14:34:28] <nygren > will do
[14:34:41] <kadukoafs@gmail.com/barnowlD26A4D0C> Thanks!
[14:34:44] Martin Duke2 joins the room
[14:36:05] yuki goto leaves the room
[14:36:09] yuki goto joins the room
[14:36:24] Bob Briscoe leaves the room
[14:36:32] Bob Briscoe joins the room
[14:36:37] <ted.h> Yeah, I'm never very happy with the "everything is already ruined" argument, but I grant that in this case there are lots of other binding possibilities.  Brian has a list of dozens.
[14:37:03] <Colin Perkins> working
[14:37:06] <Matthew Ford> yay!
[14:37:07] <Samuel Jero_> Sounding good!
[14:37:08] mcmanus joins the room
[14:37:08] <ted.h> I'm mostly worried about people doing it for simplicitly
[14:37:08] <Vicky Risk> good
[14:37:09] <nygren > did that work?
[14:37:09] <Buck> nice
[14:37:11] <kadukoafs@gmail.com/barnowlD26A4D0C> Meetecho: sounds good
[14:37:18] <Meetecho> ack
[14:37:24] <lucas_pardue> Mike Czech, who's he?
[14:37:29] <mcmanus> https://github.com/quicwg/base-drafts/issues/232 seems like a relevant place to put comments where we can remember them
[14:37:37] Simon Pietro Romano leaves the room
[14:37:40] <kadukoafs@gmail.com/barnowlD26A4D0C> Mike Czech, the promoter for the Prague meeting
[14:37:54] Simon Pietro Romano joins the room
[14:38:31] yuki goto leaves the room
[14:38:42] yuki goto joins the room
[14:39:07] <kadukoafs@gmail.com/barnowlD26A4D0C> The design is Martin Thomson; Martin will fix all our problems.
[14:44:32] <nygren > Using a few bits of the connection-id to indicate its length might be worth exploring.
[14:44:33] yuki goto leaves the room
[14:44:46] yuki goto joins the room
[14:45:48] <nygren > (as much of a pain as that will be for hardware)
[14:46:18] ekaterina leaves the room
[14:46:49] <Martin Thomson> praveen is correct; the cost of this design is that you will need to know the version
[14:47:19] <Kyle Rose> I think this design is in danger of being limited to 8 types forever, as older hardware will not transit any newer packet types
[14:47:38] <Martin Thomson> Kyle: that is a risk we're willing to take
[14:48:16] <ajs> Really?  Everything we’ve ever learned with other basic things like that tells use that such a built-in no-upgrade-path hurts later
[14:48:22] <Martin Thomson> we're going to talk later about grease for these fields so that middleboxes see almost entirely random packets
[14:48:25] <Kyle Rose> ajs: agreed
[14:48:29] Julian Reschke_3563 joins the room
[14:48:35] <ajs> We’re still living with DNS’s stupid 1980s limitations because of that kind of decision
[14:49:15] <Kyle Rose> mt: ok, that's good. I was afraid we'd abandoned the principle of hiding as much info from middleboxes as possible (or, at least, making it so the middleboxes will need to pass most packets through unmolested for basic interop to work)
[14:49:20] <Buck> who is speaking?
[14:49:24] <Martin Thomson> Andrew, we'd be happy to take that input.  Keep in mind that the sorts of things you are talking about here apply only to the short packet, which is only produced and consumed when you have state
[14:49:26] <ajs> that sounds better, yes
[14:49:41] <ajs> well, as long as the grease is applied it might be ok
[14:49:49] <nygren > Having a version in the short packets will be needed for load balancers being able to know how to route them if we ever want to change the format.  Unlike the version in TLS TCP frames, the UDP packets here otherwise lack context.
[14:49:50] <Martin Thomson> emphasis on *might* :)
[14:51:03] <ajs> it’s the Internet. The emphasis is always on “might”
[14:51:52] <Martin Thomson> the packet number size is way too large (640k should be enough for anyone)
[14:52:11] <Simon Pietro Romano> General (and long) comment on version -02 of the protocol. I do believe it is a real pity that the group decided to stop the engines on multipath. I know you will answer it is (perhaps) going to be implemented as a QUIC extension. Though, I had much appreciated the fact that multipath had been considered as a first class citizen in the protocol. BTW, the code base was already there (with some tweaking we have been able to implement a bare-bones multipath functionality into the Chromium stack). I'd be interested in knowing the reasons behind that, given the assumption that multipath is a well-recognized desired feature. I'd be interested in knowing other people's ideas on that. P.S. Disregard this if you think it is disruptive for the ongoing conversation.
[14:52:13] <Kyle Rose> Grease is so far unproven as a mechanism for defeating ossification, but we're about to find out with both QUIC and TLS 1.3
[14:52:23] <ted.h> @Martin  Some of the people in the room may not be old enough to know why that is snark.
[14:53:35] Martin Thomson leaves the room
[14:53:55] <Matthew Ford> Transport Parameter Negotiation slide
[14:54:07] <nygren > I guess as long as the server can always negotiate that the client "always send long format with the version" then the server can do that if it is behind a load balancer that needs to know the version.
[14:54:07] JoeHallCDT leaves the room
[14:54:11] Yoshiro Yoneya_5680 joins the room
[14:54:14] <kadukoafs@gmail.com/barnowlD26A4D0C> Surely they know what emacs is, though!
[14:54:47] <Matthew Ford> MTU slide
[14:54:50] <Buck> emacs maybe, but no way vi....
[14:55:11] <Kyle Rose> Am I correct in interpreting the earlier conversation as "everything prior to the final cleartext packet requires no load balancer state"?
[14:55:21] JoeHallCDT joins the room
[14:55:45] <kadukoafs@gmail.com/barnowlD26A4D0C> Kyle: I think I had a different interpretation; let me type some
things.
[14:56:28] <Matthew Ford> Packet Loss &c slide
[14:57:05] nagesh javali joins the room
[14:57:52] <kadukoafs@gmail.com/barnowlD26A4D0C> I was hearing that, if you have a load-balancer setup, it has to be
able to deal with zero or "random" connection IDs, because they will
be used for the initial cleartext packets.  Therefore, there's no
reason to try and do something fancy once the server does have a
chance to select something, so the server should just pick the
client's initial value in most cases, and the load balancer would be
based on some sort of persistent hashing-like thing.
[14:57:52] Stefan Eissing joins the room
[14:58:11] Maire Reavy joins the room
[14:58:38] <Kyle Rose> That's not going to work. That works only for consistent-hash-based load balancing, not for other mechanisms that depend on persistent state.
[14:58:39] JoeHallCDT leaves the room: Replaced by new connection
[14:58:39] JoeHallCDT joins the room
[14:58:42] Dan York leaves the room
[14:58:49] <kadukoafs@gmail.com/barnowlD26A4D0C> Now, if the server wants to do fancier load balancing to handle
varying traffic levels from different clients, it still has the option
to do so by setting a new connection ID after the handshake, that
moves the client around as needed, but that might require sending some
state around.
[14:59:57] Cedrick Mbeyet joins the room
[15:00:08] ajs leaves the room
[15:00:15] sam noman leaves the room
[15:00:26] <Kyle Rose> Requiring potentially global state sharing defeats the purpose of having a server-chosen connection ID, which is to allow for arbitrary stateless load balancing schemes. Is the existing design driven by the need to support multiple packets on the first flight? Not sure there's a solution to that problem in the stateless space.
[15:00:29] sa na joins the room
[15:00:41] Chi-Jiun Su_4101 leaves the room
[15:00:53] Cedrick Mbeyet leaves the room
[15:01:16] <nygren > I believe the property needed is that the server-proposed connection ID needs to be switched to before any server state is needed.
[15:01:27] <kadukoafs@gmail.com/barnowlD26A4D0C> I mean, it might be possible to do the server-chosen ID and remain
stateless, but I haven't absorbed the handshake process well enough
yet.
[15:01:42] <Matthew Ford> Other side
[15:01:48] <Matthew Ford> 'Other' slide
[15:02:02] <kadukoafs@gmail.com/barnowlD26A4D0C> If the server state needed is just "which versions you are willing to
negotiate" and something similar to that, that should be okay, though,
right?
[15:02:17] JoeHallCDT leaves the room
[15:02:23] <Matthew Ford> Sean Turner presenting
[15:02:25] <kadukoafs@gmail.com/barnowlD26A4D0C> Martin is getting back to his computer now, though, and will hopefully
read scrollback.
[15:02:38] <Kyle Rose> Yeah, I'll wait before responding further
[15:02:44] sa na leaves the room
[15:02:50] Martin Thomson joins the room
[15:02:56] sa no joins the room
[15:03:21] <Martin Thomson> Simon Pietro Romano: please take that to the list: jabber is not a good venue for this sort of discussion
[15:03:27] Yoshiro Yoneya_5666 joins the room
[15:03:32] Dale Worley leaves the room
[15:03:40] Dale Worley joins the room
[15:04:37] Tony Hansen joins the room
[15:04:46] <Martin Thomson> Simon Pietro Romano: it's also against our charter, so it might be disruptive: we agreed to do multipath, just not now
[15:04:47] <Simon Pietro Romano> @Martin: already done. Doesn't matter.
[15:06:00] Arash Molavi Kakhki leaves the room
[15:06:28] <Matthew Ford> Ian Swett presenting QUIC-RECOVERY
[15:07:18] <Matthew Ford> Slide 2
[15:07:18] <Matthew Ford> Slide 2
[15:07:21] <craigt> https://www.ietf.org/proceedings/98/slides/slides-98-quic-recovery-00.pdf
[15:08:20] <Buck> anybody know where the http slides are?
[15:08:46] <kadukoafs@gmail.com/barnowlD26A4D0C> https://datatracker.ietf.org/meeting/98/HTTP-QUIC.PDF ?
[15:09:01] <kadukoafs@gmail.com/barnowlD26A4D0C> Hmm, that's a sad URL, isn't it.
[15:09:13] <Buck> hrm, I'm getting not found.
[15:09:26] Tony Hansen_8856 joins the room
[15:09:35] <craigt> https://www.ietf.org/proceedings/98/slides/slides-98-quic-http-00.PDF
[15:09:51] <Buck> craigt: worked, thx!
[15:10:07] <craigt> at the tail of the main agenda; refresh if you haven't recently
[15:11:43] <Martin Thomson> https://github.com/quicwg/wg-materials/tree/master/ietf98
[15:13:52] <Lucas Clemente> MIC: quic-go currently has a wip-PR for it, will probably be merged next week
[15:14:18] <craigt> https://datatracker.ietf.org/meeting/98/session/quic <- also has discussed docs at the base
[15:15:35] <Lucas Clemente> thx :)
[15:16:51] <Buck> I'm switched names to QCRAM for disambiguation
[15:18:10] giovane.moura joins the room
[15:18:56] <Kyle Rose> Ben: I'm just going to post to the list, because this turned from a clarification question into a discussion
[15:19:18] Safiqul Islam joins the room
[15:19:36] <kadukoafs@gmail.com/barnowlD26A4D0C> Sure.
[15:20:43] Spencer Dawkins leaves the room
[15:20:58] Spencer Dawkins joins the room
[15:21:08] <Matthew Ford> Handshake protocol selection #45
[15:21:39] ajs joins the room
[15:21:52] ajs leaves the room
[15:23:57] Kyle Rose leaves the room
[15:24:33] Kyle Rose joins the room
[15:25:26] Okke Timm leaves the room
[15:27:14] Okke Timm joins the room
[15:27:34] Ivaylo Petrov joins the room
[15:28:02] Matthew Ford leaves the room: Replaced by new connection
[15:28:08] Safiqul Islam leaves the room
[15:28:10] Matthew Ford joins the room
[15:30:34] Aaron Zauner joins the room
[15:30:58] Ivaylo Petrov leaves the room
[15:31:06] Matthew Ford has set the subject to: 'QUIC IETF98 WG Meeting 20170330'
[15:31:15] Ivaylo Petrov joins the room
[15:33:57] Dan York joins the room
[15:34:00] <ted.h> Or MTFC, as we call it informally
[15:34:01] Ben Niven-Jenkins joins the room
[15:35:50] yuki goto leaves the room
[15:35:52] Daniel Stenberg_1565 joins the room
[15:35:58] Bob Briscoe leaves the room
[15:36:11] yuki goto joins the room
[15:36:51] Ben Niven-Jenkins_2091 joins the room
[15:37:03] <Lucas Clemente> i feel like we are rat-holeing
[15:38:35] Peter Thatcher joins the room
[15:39:49] gryning joins the room
[15:40:02] craigt leaves the room: Disconnected: closed
[15:40:04] Yoshiro Yoneya_5666 leaves the room
[15:40:09] giovane.moura leaves the room
[15:40:15] Yoshiro Yoneya_4887 joins the room
[15:40:24] nygren leaves the room: Disconnected: No route to host
[15:40:58] <Matthew Ford> Issue 61 Slient close
[15:43:00] Ben Jenkins joins the room
[15:43:44] Spencer Dawkins leaves the room
[15:43:50] Spencer Dawkins joins the room
[15:44:32] Stefan Eissing_4660 joins the room
[15:48:19] <Meetecho> note to chairs: Martin Duke in virtual queue
[15:49:29] Aaron Falk leaves the room
[15:49:30] Aaron Falk joins the room
[15:50:56] <Olivier Bonaventure> I disagree with Jana, having a close packet is important for all intermediate devices.
[15:51:25] <Martin Thomson> ian is right about the two separable issues
[15:52:44] <Martin Thomson> mobility and NAT rebinding means that missing end-of-connection signals are going to happen less and less
[15:52:51] <ted.h> The trade-off is that one packet can be a lot of radio resources/battery power over time.
[15:53:08] <kadukoafs@gmail.com/barnowlD26A4D0C> Martin: er, will happen more and more?
[15:53:10] Sean Turner joins the room
[15:53:37] <Stefan Eissing_4660> Can one fake end-of-connection?
[15:54:28] John Border joins the room
[15:54:45] <Olivier Bonaventure> Compare a long-lived connection on smartphone. With TCP, SYN at the beginning, data exchange at any time and then FIN at the end. With QUIC, we should not need to send keep alive messages to ensure that state remains in NATs and other middleboxes. Otherwise, we’ll drain battery on many mobile devices.
[15:56:30] Matthew Ford leaves the room
[15:56:42] Matthew Ford joins the room
[15:57:44] <Olivier Bonaventure> Totally agree with Stuart. If base case is explicitly signalled terminations, then middleboxes can be optimised for that as with TCP today.
[15:58:15] Fan Yang leaves the room
[15:58:21] <Colin Perkins> +1 to Gorry – explicit close is very important
[15:59:53] <Martin Thomson> this issue got seriously derailed
[15:59:57] <ted.h> @Olivier  If the state is such that the client will initiate the next request, you don't need keep alives, as the connection id survives the nat rebinding.
[16:00:23] <Martin Thomson> but it's good that we're discussing this issue
[16:02:11] nygren joins the room
[16:02:13] <Matthew Ford> QUIC Unprotected packets
[16:03:45] <nygren > [now that my network is back]  It seems like we were conflating at least four independent but related topics in the last discussion:  
[16:03:48] <nygren >   1) whether silent close is available as an option for optimizing mobile end-device power
  2) whether explicit transport connection close is provided as a transport by QUIC.    
  3) whether transport connection close can be made visible to just endpoint and/or middleboxes.
      - as an optimization to allow NATs to prune (majority of?) bindings
      - also valuable for servers for dropping state      
  4) application-level soft-close semantics
       (force all apps to re-invent, or have a TAPS service/API that QUIC implements?)
[16:03:57] <Martin Thomson> safari
[16:06:37] Yoshiro Yoneya_4887 leaves the room
[16:06:53] Yoshiro Yoneya_5697 joins the room
[16:08:10] each joins the room
[16:08:18] Dan York leaves the room
[16:08:26] each leaves the room
[16:09:45] Sean Turner leaves the room: Replaced by new connection
[16:09:49] Sean Turner joins the room
[16:09:49] Matthew Ford leaves the room
[16:10:29] Sean Turner leaves the room
[16:10:37] Matthew Ford joins the room
[16:10:43] <Lucas Clemente> with option 3 a client cannot choose to only support chacha, but always has to implement aes.
[16:11:06] <ted.h> @Lucas Do you want that reflected to the room?
[16:11:26] <Lucas Clemente> if there's time ;)
[16:11:36] <kadukoafs@gmail.com/barnowlD26A4D0C> I thought AES-GCM was MTI for TLS 1.3
[16:12:00] dkg leaves the room
[16:12:02] <Martin Duke> Can we take a hum?
[16:12:03] dkg joins the room
[16:12:29] <kadukoafs@gmail.com/barnowlD26A4D0C>    In the absence of an application profile standard specifying
   otherwise, a TLS-compliant application MUST implement the
   TLS_AES_128_GCM_SHA256 cipher suite and SHOULD implement the
   TLS_AES_256_GCM_SHA384 and TLS_CHACHA20_POLY1305_SHA256 cipher
   suites. (see Appendix B.4)
[16:13:10] <Lucas Clemente> hums for B
[16:13:16] Sean Turner joins the room
[16:13:22] <Martin Duke> Hum for FNV-1a
[16:13:22] <Buck> hums for B
[16:13:24] <Samuel Jero_> hums for B
[16:14:20] <nygren > mnot didn't even need to flip a coin!
[16:15:11] koji leaves the room
[16:15:30] <Matthew Ford> Moving to issue 391
[16:16:05] Boo Radley joins the room
[16:16:15] koji joins the room
[16:21:19] koji leaves the room
[16:21:52] <Buck> CDN vs carrieres is very different, because CDNs have certs and transparent middleboxes don't, no?
[16:22:10] <ted.h> Actually that suggests letting the server initiate this and then the client echo might be useful, if you are not using the packet number
[16:22:22] lucas_pardue leaves the room
[16:22:49] <ted.h> So, the server sends "token" and then the client decides whether to echo "token".
[16:23:33] <ted.h> If the server does this once per RTT, the network can see something that indicates the server's view of RTT even if the token isn't echoed
[16:23:34] Aaron Falk leaves the room
[16:25:27] <Buck> not clear why e2e RTT vs bottlneck segment is materially different.   it should be possible to normalize e2e RTT.
[16:25:34] nygren leaves the room
[16:26:22] nygren joins the room
[16:27:26] <Kyle Rose> I wonder if Jim's proposal is a better approach for something PLUS-like?
[16:27:46] <Marten Seemann> I like ted.h's proposal
[16:28:43] Matthew Ford leaves the room
[16:28:54] Matthew Ford joins the room
[16:29:05] <Buck> It's a style of AQM, i.e. where there is a separation between measurment and enforcement.   If you go classic AQM, where the estimation and the markign are in the same box, there's not as much of an issue.
[16:29:43] Kyle Rose leaves the room
[16:30:19] <ted.h> Ian points out that what I suggested is basically active probing, and that you could do that without tying it to QUIC (Ben's point as well).  I agree, but there is a chance that a particular protocol has a different path or character, so I don't think it is that bad to do it within QUIC.
[16:30:30] Valery Smyslov leaves the room
[16:30:44] ekaterina joins the room
[16:30:53] kadukoafs@gmail.com/barnowlD26A4D0C leaves the room
[16:30:53] svaldez@tls.ctf.network leaves the room
[16:30:54] <nygren > If clients and servers need to mutually agree to send this, and this is explicitly not always sent, then it seems like there can be net value.
[16:30:59] Matthew Ford leaves the room
[16:31:01] nygren leaves the room
[16:31:06] ekaterina leaves the room
[16:31:10] Simon Pietro Romano leaves the room
[16:31:15] mcmanus leaves the room
[16:31:18] <ted.h> @nygren, I think it is at most once per RTT.
[16:31:21] Kaoru Maeda leaves the room
[16:31:35] Francis Teague leaves the room
[16:31:49] sa no leaves the room
[16:32:36] Magnus Westerlund leaves the room
[16:32:48] Ben Niven-Jenkins leaves the room
[16:32:48] Tony Hansen_8856 leaves the room
[16:32:48] Okke Timm leaves the room
[16:32:48] Edward Lopez leaves the room
[16:32:48] Ronald van der Pol leaves the room
[16:32:48] Yoshiro Yoneya_5697 leaves the room
[16:32:48] Colin Perkins leaves the room
[16:32:49] Iheanyi Obasi leaves the room
[16:32:49] Julian Reschke leaves the room
[16:32:49] Maire Reavy leaves the room
[16:32:49] Daniel Stenberg leaves the room
[16:32:49] Olivier Bonaventure leaves the room
[16:32:49] Tony Hansen leaves the room
[16:32:49] Wendy Seltzer leaves the room
[16:32:49] Alain Durand_7025 leaves the room
[16:32:49] Aaron Zauner leaves the room
[16:32:49] Ivaylo Petrov leaves the room
[16:32:49] Julian Reschke_3563 leaves the room
[16:32:49] Daniel Stenberg_1565 leaves the room
[16:32:49] yuki goto leaves the room
[16:32:49] Ken Murchison leaves the room
[16:32:49] Randell Jesup leaves the room
[16:32:49] Stefan Eissing_4660 leaves the room
[16:32:49] nagesh javali leaves the room
[16:32:49] cameron byrne leaves the room
[16:32:49] Peter Thatcher leaves the room
[16:32:49] Charles Krasic leaves the room
[16:32:49] Martin Duke2 leaves the room
[16:32:49] Martin Duke leaves the room
[16:32:49] Samuel Jero_ leaves the room
[16:32:49] Yoshiro Yoneya_5680 leaves the room
[16:32:49] julien maisonneuve leaves the room
[16:32:49] a a leaves the room
[16:32:49] Spencer Dawkins leaves the room
[16:32:49] Stanislav Slusny leaves the room
[16:32:49] John Border leaves the room
[16:32:49] Alain Durand_3681 leaves the room
[16:32:49] Safa Almalki leaves the room
[16:32:49] Jim Welch leaves the room
[16:32:49] Dapeng Liu leaves the room
[16:32:49] Alain Durand leaves the room
[16:32:49] s d leaves the room
[16:32:49] p m leaves the room
[16:32:49] Padma Bhooma leaves the room
[16:32:49] Marten Seemann leaves the room
[16:32:49] Boo Radley leaves the room
[16:32:49] Vicky Risk leaves the room
[16:32:49] John Leslie leaves the room
[16:32:49] Richard Bradbury leaves the room
[16:32:49] Ben Niven-Jenkins_2091 leaves the room
[16:32:49] Satoru Kanno leaves the room
[16:32:49] Natasha Rooney leaves the room
[16:32:49] Stefan Eissing leaves the room
[16:32:49] Lucas Clemente leaves the room
[16:32:49] Ben Jenkins leaves the room
[16:32:49] Yang Song leaves the room
[16:32:49] Dale Worley leaves the room
[16:33:11] Yoshiro Yoneya leaves the room
[16:33:17] Martin Thomson leaves the room
[16:34:50] Yoshiro Yoneya joins the room
[16:35:34] koji joins the room
[16:35:47] koji leaves the room
[16:35:49] gryning leaves the room: Disconnected: timeout during writing
[16:36:04] ted.h leaves the room: Replaced by new connection
[16:36:04] ted.h joins the room
[16:38:14] Samuel Jero leaves the room
[16:40:20] ted.h leaves the room
[16:40:47] Meetecho leaves the room
[16:42:07] ekaterina joins the room
[16:42:11] frodek leaves the room
[16:46:46] Buck leaves the room
[16:47:30] Sean Turner leaves the room
[16:49:41] dkg leaves the room
[16:51:55] ekaterina leaves the room
[16:53:47] svaldez@tls.ctf.network joins the room
[16:55:04] g.e.montenegro leaves the room
[16:55:48] svaldez@tls.ctf.network leaves the room
[16:58:37] JoeHallCDT joins the room
[16:58:47] JoeHallCDT leaves the room
[16:59:41] Yoshiro Yoneya joins the room
[16:59:41] Yoshiro Yoneya leaves the room
[17:00:17] Yoshiro Yoneya leaves the room
[17:24:09] g.e.montenegro joins the room
[17:24:38] g.e.montenegro leaves the room
[17:46:35] giovane.moura joins the room
[18:01:38] gryning joins the room
[18:02:14] giovane.moura leaves the room
[18:10:05] nygren joins the room
[18:10:50] gryning leaves the room: Disconnected: closed
[18:18:04] frodek joins the room
[18:25:22] nygren leaves the room
[18:26:32] frodek leaves the room
[18:28:46] nygren joins the room
[18:33:08] nygren leaves the room
[18:34:57] nygren joins the room
[19:06:25] nygren leaves the room
[19:14:31] nygren joins the room
[19:39:36] svaldez@tls.ctf.network joins the room
[19:55:46] nygren leaves the room
[19:56:22] svaldez@tls.ctf.network leaves the room
[20:06:17] nygren joins the room
[20:08:27] Sean Turner joins the room
[20:11:15] Sean Turner leaves the room: Replaced by new connection
[20:11:18] Sean Turner joins the room
[20:15:04] Sean Turner leaves the room
[20:21:31] nygren leaves the room
[20:36:29] nygren joins the room
[20:45:09] nygren leaves the room
[20:52:42] nygren joins the room
[21:42:20] nygren leaves the room
[22:52:44] nygren joins the room
[23:20:53] ekaterina joins the room
[23:35:47] nygren leaves the room
[23:42:53] ekaterina leaves the room
Powered by ejabberd - robust, scalable and extensible XMPP server Powered by Erlang Valid XHTML 1.0 Transitional Valid CSS!