The built-in desks for each seat are really grand. That should be a mtgvenue standard.
", "time": "2024-11-06T09:34:02Z"}, {"author": "Victor Pascual", "text": "[40 min - Open issues, updates to multipath QUIC. slides]
\ngithub link to slides seems to be broken (404)
\nshould be https://github.com/quicwg/wg-materials/blob/main/ietf121/multipath.pdf
", "time": "2024-11-06T09:37:59Z"}, {"author": "Lucas Pardue", "text": "thanks Victor
", "time": "2024-11-06T09:42:22Z"}, {"author": "Alan Frindell", "text": "+1 to scheduling virtual interop for reliable reset before bangkok
", "time": "2024-11-06T09:43:46Z"}, {"author": "Lucas Pardue", "text": "agenda should have a fixed link now -
", "time": "2024-11-06T09:45:12Z"}, {"author": "Kazuho Oku", "text": "I think two things are separate: a) if ack-frequency is good enough as a signalling scheme, and b) how it should be integrated to CC.
", "time": "2024-11-06T09:49:25Z"}, {"author": "Kazuho Oku", "text": "The latter is difficult I agree, but the signalling scheme is designed to handle various conditions that we might need in tuning b.
", "time": "2024-11-06T09:50:17Z"}, {"author": "Lucas Pardue", "text": "@meetecho any chance we refresh the slides for the next presentation during this session please?
", "time": "2024-11-06T09:52:31Z"}, {"author": "Lorenzo Miniero", "text": "@Lucas Pardue if they were updated on the datatracker, they should have been pulled automatically in Meetecho too. If it's QLOG, we already see a revision 01 for instance from 9.40am
", "time": "2024-11-06T09:53:37Z"}, {"author": "Lucas Pardue", "text": "thats the one, thank you!
", "time": "2024-11-06T09:53:48Z"}, {"author": "Martin Duke", "text": "There are two things relating to ACK_FREQUENCY maturity. 1) We should encourage extremely wide deployment of support for honoring these frames and (2) most deployments should be conservative about reducing the peer's ack frequency until they understand the impact on their congestion control
", "time": "2024-11-06T09:56:30Z"}, {"author": "Guoye Zhang", "text": "Should it be PATHS_BLOCKED instead of PATH_BLOCKED?
", "time": "2024-11-06T09:58:37Z"}, {"author": "Jonathan Lennox", "text": "Guoye: I think you're right - I just commented that on the PR for #463
", "time": "2024-11-06T09:59:26Z"}, {"author": "Eric Kinnear", "text": "Christian: Great point about sending them in the same packet, that covers everything I was looking for
", "time": "2024-11-06T10:00:11Z"}, {"author": "Martin Thomson", "text": "time check for this agenda slot
", "time": "2024-11-06T10:22:15Z"}, {"author": "Kazuho Oku", "text": "The issue Mike descibed exists with RFC 9000 running out of CIDs too?
", "time": "2024-11-06T10:23:58Z"}, {"author": "Kazuho Oku", "text": "So I think we are fine (in sense that MPQUIC is no worse than RFC 9000).
", "time": "2024-11-06T10:24:20Z"}, {"author": "Magnus Westerlund", "text": "The point I was going to make is that if we remove path migration then we also have to decided if path migration is forbidden and if that results in a protocol violation.
", "time": "2024-11-06T10:25:07Z"}, {"author": "Martin Duke", "text": "+1 to Lars
", "time": "2024-11-06T10:29:23Z"}, {"author": "Martin Thomson", "text": "Lars makes a very good point. See also David's points about reliable/partial reset and ack frequency.
", "time": "2024-11-06T10:30:37Z"}, {"author": "Robin Marx", "text": "I'm indeed here :)
", "time": "2024-11-06T10:36:53Z"}, {"author": "Robin Marx", "text": "nope
", "time": "2024-11-06T10:40:33Z"}, {"author": "Lars Eggert", "text": "I don't care what we do as long as there is a Rust crate
", "time": "2024-11-06T10:55:07Z"}, {"author": "Martin Thomson", "text": "xargs -0 might make a case for nul-byte delimiters instead.
", "time": "2024-11-06T10:55:35Z"}, {"author": "Lars Eggert", "text": "Did someone say CBOR?
", "time": "2024-11-06T10:56:01Z"}, {"author": "Lars Eggert", "text": ":grinning:
", "time": "2024-11-06T10:56:09Z"}, {"author": "Kazuho Oku", "text": "Would love to see a poll on which options people dislike :-)
", "time": "2024-11-06T10:56:33Z"}, {"author": "Martin Thomson", "text": "Can a string literal in JSON include a literal CR or LF?
", "time": "2024-11-06T10:56:41Z"}, {"author": "Kazuho Oku", "text": "CR LF MUST be escaped
", "time": "2024-11-06T10:58:23Z"}, {"author": "Martin Thomson", "text": "(answering my own question) No, JSON cannot include CR or LF, except in the whitespace, which is ignored
", "time": "2024-11-06T10:58:25Z"}, {"author": "Martin Thomson", "text": "echo '[';sed -e 'y/\\r\\n/,,/';s/,,/,/g' \"$f\"; echo ']'
", "time": "2024-11-06T11:04:13Z"}, {"author": "Christian Huitema", "text": "It is not just tools -- we want to use text based tools for parsing the logs!
", "time": "2024-11-06T11:04:33Z"}, {"author": "Lars Eggert", "text": "I think as long as we pick something that has a spec maintained by a reasonable org that is also in wider use, not going with an IETF spec is a very defensible choice
", "time": "2024-11-06T11:05:26Z"}, {"author": "Martin Thomson", "text": "echo '[';sed -e $'y/\\x1e/,/' \"$f\"; echo ']'
", "time": "2024-11-06T11:05:35Z"}, {"author": "Lars Eggert", "text": "Want me to do perl?
", "time": "2024-11-06T11:05:54Z"}, {"author": "Martin Thomson", "text": "The two seem pretty similar to me.
", "time": "2024-11-06T11:05:55Z"}, {"author": "Martin Thomson", "text": "No one should ever do perl
", "time": "2024-11-06T11:06:35Z"}, {"author": "Robin Marx", "text": "@Lars: that's good feedback, thanks
", "time": "2024-11-06T11:06:49Z"}, {"author": "Matt Joras", "text": "Lars: We don't really have even a single alternative, there's kind of two competing ones and neither is really maintained by an \"org\".
\njsonlines is maintained by a guy.
\nndjson I'm not sure is even claimed to be owned? But Lucas keep me honest here
", "time": "2024-11-06T11:07:34Z"}, {"author": "Martin Thomson", "text": "FWIW, the above transforms work in both directions fairly trivially.
", "time": "2024-11-06T11:08:31Z"}, {"author": "Lucas Pardue", "text": "On the issues I linked in the slides, there seemed to be an interest in aligning the two and maybe even defining the media types somehow.
", "time": "2024-11-06T11:08:55Z"}, {"author": "Lucas Pardue", "text": "Issues on ndjson or json lines repo ^^
", "time": "2024-11-06T11:09:15Z"}, {"author": "Martin Thomson", "text": "I'm glad that we don't have the old IETF. Because discussions of a PUNCH_ME_NOW frame in the olden days might have had bad outcomes.
", "time": "2024-11-06T11:09:41Z"}, {"author": "Lucas Pardue", "text": "My concern is, the JSON-SEQ process shook out something in the design of the format that is good for something we are overlooking. I didn't go into the archaeology of JSON-SEQ process and it might be helpful for people involved in that to weigh in
", "time": "2024-11-06T11:10:47Z"}, {"author": "Jonathan Lennox", "text": "I think one of the main things is that the thing encapsulated in JSON-SEQ is 100% standard JSON and not \"JSON except you can't use newlines\" which is poorly-defined.
", "time": "2024-11-06T11:12:11Z"}, {"author": "Martin Thomson", "text": "That STUN argument is not a good one. Is there a better argument for using QUIC? It seems like there might be.
", "time": "2024-11-06T11:12:14Z"}, {"author": "Martin Thomson", "text": "\"That STUN argument\" refers to the argument that Marten used to argue against using STUN.
", "time": "2024-11-06T11:13:28Z"}, {"author": "Christian Huitema", "text": "@Martin I like that we can have address discovery potentially with ANY server. It is good for P2P, because of \"don't stick out\". But it can also be used for debug, etc.
", "time": "2024-11-06T11:13:53Z"}, {"author": "Martin Thomson", "text": "Christian: that's a good one.
", "time": "2024-11-06T11:14:16Z"}, {"author": "Martin Thomson", "text": "I think that the authentication in STUN is also particularly complex for this purpose.
", "time": "2024-11-06T11:14:34Z"}, {"author": "Martin Thomson", "text": "STUN relies on shared keys, which would mean the use of exporters or something like that.
", "time": "2024-11-06T11:14:56Z"}, {"author": "Christian Huitema", "text": "Yes, having everything encrypted is a also a huge plus.
", "time": "2024-11-06T11:15:29Z"}, {"author": "Jonathan Lennox", "text": "Generally the address discovery part of STUN is unauthenticated (since the cost of sending an error message is comparable to the cost of replying)
", "time": "2024-11-06T11:15:33Z"}, {"author": "Jonathan Lennox", "text": "And it's never encrypted
", "time": "2024-11-06T11:15:41Z"}, {"author": "Martin Thomson", "text": "Jonathan, I'm talking now about the consent-y bits.
", "time": "2024-11-06T11:15:57Z"}, {"author": "Martin Thomson", "text": "32 + 16 > 128 !
", "time": "2024-11-06T11:17:19Z"}, {"author": "Robin Marx", "text": "@Jonathan Lennox: not sure you get that much of benefit of JSON-SEQ records being \"100% standard JSON\" in practice though...
", "time": "2024-11-06T11:17:22Z"}, {"author": "Jonathan Lennox", "text": "Robin: mostly that you can safely take the output of any tool that generates JSON and put it in JSON-SEQ, without having to postprocess it, I think
", "time": "2024-11-06T11:18:05Z"}, {"author": "Martin Thomson", "text": "That is correct
", "time": "2024-11-06T11:19:05Z"}, {"author": "Robin Marx", "text": "Jonathan: yeah sure, in a general sense, but I don't see when that would be a main use case for qlog specifically
", "time": "2024-11-06T11:19:06Z"}, {"author": "Ted Hardie", "text": "They closed the queue, but this makes me wonder if it would be useful to include the connection ID associated with the newly observed IP address in the OBSERVED_ADDRESS frame.
", "time": "2024-11-06T11:21:37Z"}, {"author": "Christian Huitema", "text": "If not a WG item, then what? Private draft? Reserve a code point with IANA?
", "time": "2024-11-06T11:22:44Z"}, {"author": "Valentin Go\u0219u", "text": "I think the OBSERVED_ADDRESS frame would be useful outsde the P2P use case
", "time": "2024-11-06T11:22:54Z"}, {"author": "Martin Thomson", "text": "Nothing stopping you from deploying this.
", "time": "2024-11-06T11:23:46Z"}, {"author": "Lars Eggert", "text": "A use case would be to eventually get WebRTC over QUIC. Because I would love to get rid of that pile of protocol spaghetti.
", "time": "2024-11-06T11:24:12Z"}, {"author": "Martin Thomson", "text": "But this doesn't seem particularly clear in terms of design and use cases. And even if there were use cases, it might be that the time of those in the group is better spent elsewhere. Lots of things to consider ahead of adoption.
", "time": "2024-11-06T11:24:39Z"}, {"author": "Matt Joras", "text": "Lars why would anyone do WebRTC over QUIC and not just straight MoQ?
", "time": "2024-11-06T11:25:08Z"}, {"author": "Martin Thomson", "text": "WebRTC over QUIC depends on many things, but it is quite likely that there is an intermediate state that uses STUN and ICE.
", "time": "2024-11-06T11:25:11Z"}, {"author": "Jonathan Lennox", "text": "Robin: I guess mostly it means you have to make sure your JSON tool has a \"no pretty-printing\" option and it's turned on.
", "time": "2024-11-06T11:25:18Z"}, {"author": "Martin Thomson", "text": "WebRTC over QUIC might be MoQ at some point in that evolution (incl. the end point)
", "time": "2024-11-06T11:25:45Z"}, {"author": "Lars Eggert", "text": "@Matt: I thought MOQ was the path here, but people this week told me that WebRTC wasn't of interest to MOQ. (Maybe I was talking to thr wrong people?)
", "time": "2024-11-06T11:26:10Z"}, {"author": "Lucas Pardue", "text": "\n\nIf not a WG item, then what? Private draft? Reserve a code point with IANA?
\n
We do have the processes in place to do this and reserve codepoints for interop, without needing standards action.
\nNo strong opinion on this, just highlighting its possible
", "time": "2024-11-06T11:26:21Z"}, {"author": "Matt Joras", "text": "Lars: Basically the conference/VC use case is in the charter for MoQ
", "time": "2024-11-06T11:26:40Z"}, {"author": "Jonathan Lennox", "text": "I think there's still some uncertainty about whether MoQ can actually get its latency as low as RTP, though it can certainly get close.
", "time": "2024-11-06T11:27:34Z"}, {"author": "Matt Joras", "text": "WebRTC basically exists for conferencing and then live video/gaming/whatever.
", "time": "2024-11-06T11:27:34Z"}, {"author": "Lars Eggert", "text": "Glad to hear. So maybe it's just that the current focus is elsewhere for MOQ.
", "time": "2024-11-06T11:27:37Z"}, {"author": "Christian Huitema", "text": "I would love to to P2P MoQ!
", "time": "2024-11-06T11:27:44Z"}, {"author": "Jonathan Lennox", "text": "Out of scope for here, but does MoQ support bidirectional links (i.e. both sides are both publishers and subscribers)?
", "time": "2024-11-06T11:28:49Z"}, {"author": "Lars Eggert", "text": "@Christian You mean picoquic hasn't implemented that already?? Shocked.
", "time": "2024-11-06T11:28:51Z"}, {"author": "Matt Joras", "text": "I wouldn't say the current MoQ focus is elsewhere, it just hasn't gotten much into the higher level details than the \"transport\" as MoQ has defined it
", "time": "2024-11-06T11:28:52Z"}, {"author": "Christian Huitema", "text": "@Lars LOL. But implementing MoQ is a bit like the dof running after the car...
", "time": "2024-11-06T11:29:30Z"}, {"author": "Christian Huitema", "text": "(dog)
", "time": "2024-11-06T11:29:42Z"}, {"author": "Matt Joras", "text": "Jonathan: yes I believe so, at least that's the intention
", "time": "2024-11-06T11:29:45Z"}, {"author": "Tam\u00e1s L\u00e9vai", "text": "@Jonathan: an example is the relay--relay connection
", "time": "2024-11-06T11:31:30Z"}]