[{"author": "Alejandro Sede\u00f1o", "text": "

New topic, who dis?

", "time": "2023-11-06T14:35:33Z"}, {"author": "David Schinazi", "text": "

If anyone in this chat would like something said at the microphone, please say it here prefixed with \"mic:\" and I will repeat it

", "time": "2023-11-06T14:38:20Z"}, {"author": "Lucas Pardue", "text": "

does the atomic tool reside in a lead-lined toolbox?

", "time": "2023-11-06T14:49:30Z"}, {"author": "Mike English", "text": "

Where do I route beers for more green on that chart? Eric?

", "time": "2023-11-06T14:52:08Z"}, {"author": "Lucas Pardue", "text": "

send cake instead - https://www.howtogeek.com/817409/mozilla-sent-microsoft-a-cake-for-internet-explorers-death/

", "time": "2023-11-06T14:53:02Z"}, {"author": "David Schinazi", "text": "

:cake::birthday:

", "time": "2023-11-06T14:53:55Z"}, {"author": "David Schinazi", "text": "

(huh I sent a cake emoji and now I'm seeing a baloon emoji, that's odd)

", "time": "2023-11-06T14:54:26Z"}, {"author": "Lucas Pardue", "text": "

QUIC always recommends pacing ;)

\n

https://datatracker.ietf.org/doc/html/rfc9002#name-pacing

", "time": "2023-11-06T14:56:17Z"}, {"author": "Kirill Pugin", "text": "

is there equivalent of receive rate?

", "time": "2023-11-06T14:57:24Z"}, {"author": "Luke Curley", "text": "

@kirill not with TCP algorithms, your application will have to send it to the receiver

", "time": "2023-11-06T15:05:18Z"}, {"author": "Kirill Pugin", "text": "

yeah, I guess what I meant is that there at least 2 usecases:

\n

1) I am encoder and need to know how to configure encoder to try to avoid congestion

\n

2) I am receiver - I want to know what is highest quality (for video usecases) I can download to avoid stalling at playback time

", "time": "2023-11-06T15:06:49Z"}, {"author": "Luke Curley", "text": "

yeah, ultimately the sender's congestion control algorithm is the upper bound of what can be sent over the network

", "time": "2023-11-06T15:08:22Z"}, {"author": "Kirill Pugin", "text": "

yup, so that API would tell me #1 or #2 from my example? :D

", "time": "2023-11-06T15:09:07Z"}, {"author": "Luke Curley", "text": "

the encoder and receiver should be less than this value, so it's ideal to pull it directly from the CC

", "time": "2023-11-06T15:09:59Z"}, {"author": "Luke Curley", "text": "

so yeah the encoder could call webtransport.getStats() or whatever to periodically update the encoder bitrate

", "time": "2023-11-06T15:10:40Z"}, {"author": "Erik Nygren", "text": "

We should make sure we don't run into the same attack problems that H/2 is experiencing with \"Rapid Resets\" and/or introduce a new vector. (eg, servers need to be able to bound stream limits in a way that can't be bypassed in a way similar to Rapid Resets)

", "time": "2023-11-06T15:10:44Z"}, {"author": "Luke Curley", "text": "

and the sender could periodically send the estimated bitrate to the receiver for ABR purposes

", "time": "2023-11-06T15:10:59Z"}, {"author": "Luke Curley", "text": "

pls don't try to guess the estimated bitrate on the receiver side like HLS/DASH

", "time": "2023-11-06T15:11:17Z"}, {"author": "Kirill Pugin", "text": "

@Luke, yup it would be nice to have these as 2 APIs ;)

", "time": "2023-11-06T15:11:43Z"}, {"author": "Luke Curley", "text": "

yeah I think it makes sense in MoQ :D

", "time": "2023-11-06T15:12:19Z"}, {"author": "Luke Curley", "text": "

just gotta factor in relays somehow

", "time": "2023-11-06T15:12:32Z"}, {"author": "Luke Curley", "text": "

estimated send rate is a hop-by-hop property, while stuff like ABR is end-to-end

", "time": "2023-11-06T15:12:56Z"}, {"author": "Lucas Pardue", "text": "

what document is the capsule defined in?

", "time": "2023-11-06T15:13:11Z"}, {"author": "Sauli Kiviranta", "text": "

+1 for direct from cc and sharing estimated bitrate to receiver
\nThe less decisions on derived data are made on behalf of what we think use case should need or could need the better

", "time": "2023-11-06T15:13:25Z"}, {"author": "David Schinazi", "text": "

want me to say that at the mic Lucas?

", "time": "2023-11-06T15:13:35Z"}, {"author": "Lucas Pardue", "text": "

too late :), I'll take an action to follow up

", "time": "2023-11-06T15:14:36Z"}, {"author": "Alan Frindell", "text": "

Oh I want tridirectional streams now

", "time": "2023-11-06T15:15:12Z"}, {"author": "David Schinazi", "text": "

And datapounds

", "time": "2023-11-06T15:15:29Z"}, {"author": "Kirill Pugin", "text": "

let's stay in metric system

", "time": "2023-11-06T15:15:52Z"}, {"author": "Erik Nygren", "text": "

See https://martinthomson.github.io/h2-stream-limits/draft-thomson-httpbis-h2-stream-limits.html for @Martin Thomson 's proposal for making H2 stream limits consistent with H3.

", "time": "2023-11-06T15:22:58Z"}, {"author": "Eric Gorbaty", "text": "

Commenting on the earlier WT_MAX_STREAMS thing: I'm partial to putting explicit examples in the doc to make usage for these clear.

", "time": "2023-11-06T15:23:00Z"}, {"author": "David Schinazi", "text": "

h3-over-webtransport coming up next

", "time": "2023-11-06T15:29:58Z"}, {"author": "Alan Frindell", "text": "

Are there other TLS -ish things can you do with raw quic that someone is going to ask for over WebTransport?

", "time": "2023-11-06T15:30:23Z"}, {"author": "Alan Frindell", "text": "

Oh I want to call it \"Not-ALPN\" now

", "time": "2023-11-06T15:31:41Z"}, {"author": "Luke Curley", "text": "

I like Not-ALPN

", "time": "2023-11-06T15:32:05Z"}, {"author": "Mike Bishop", "text": "

Did David just say \"no one strongly objects\" is consensus?

", "time": "2023-11-06T15:34:53Z"}, {"author": "Erik Nygren", "text": "

Do we need to define how WebTransport interacts with SVCB / HTTPS DNS RRs? (Especially since there we define a SvcParam for ALPN and if we use ALPN here then we need to define how they interact.)

", "time": "2023-11-06T15:34:56Z"}, {"author": "Luke Curley", "text": "

+1 Magnus

", "time": "2023-11-06T15:35:10Z"}, {"author": "Lucas Pardue", "text": "

WTSPN

", "time": "2023-11-06T15:35:23Z"}, {"author": "Luke Curley", "text": "

BLPN

", "time": "2023-11-06T15:35:27Z"}, {"author": "Victor Vasiliev", "text": "

My alternative proposal would be WebTransport-Subprotocols that would be an sf-list of sf-bytestrings

", "time": "2023-11-06T15:35:28Z"}, {"author": "Luke Curley", "text": "

ALPN++

", "time": "2023-11-06T15:35:32Z"}, {"author": "Jonathan Lennox", "text": "

If we call it subprotocol does that cause too much confusion with websockets?

", "time": "2023-11-06T15:35:36Z"}, {"author": "Jonathan Lennox", "text": "

(I think it shouldn't so long as it's clear.)

", "time": "2023-11-06T15:37:51Z"}, {"author": "Lucas Pardue", "text": "

+1 to not making server wait on client settings in the final version of WT

", "time": "2023-11-06T15:56:52Z"}, {"author": "Victor Vasiliev", "text": "

People are welcome to object to the proposal in question, but then they're left to implement multiple WebTransport version support in some other alternative way...

", "time": "2023-11-06T15:58:54Z"}, {"author": "Luke Curley", "text": "

put the WebTransport version in the CONNECT header? :p

", "time": "2023-11-06T15:59:38Z"}]