httpbis - Tuesday, May 19, 2020

[12:59:54] Martin Thomson joins the room
[13:00:27] Mark Nottingham joins the room
[13:00:42] Martin Thomson has set the subject to: https://github.com/httpwg/wg-materials/blob/gh-pages/interim-20-05/agenda.md
https://etherpad.ietf.org:9009/p/notes-ietf-interim-2020-httpbis-01-httpbis?useMonospaceFont=true
[13:00:53] Magnus Westerlund joins the room
[13:01:06] <Martin Thomson> At least webex is offering me the option to join from a browser this time.
[13:01:08] lpardue joins the room
[13:01:35] <Martin Thomson> Gah, video
[13:01:42] Tommy Pauly joins the room
[13:01:42] Tommy Pauly leaves the room
[13:02:33] https://etherpad.ietf.org:9009/p/notes-ietf-interim-2020-httpbis-01-httpbis?useMonospaceFont=true
[13:02:47] ChrisBox joins the room
[13:02:48] BDC joins the room
[13:04:16] BDC leaves the room
[13:04:38] bdc joins the room
[13:05:31] tfpauly joins the room
[13:05:43] mcr joins the room
[13:07:05] James Gruessing joins the room
[13:07:41] ChrisBox leaves the room
[13:09:14] ChrisBox joins the room
[13:09:31] ekr@jabber.org joins the room
[13:09:45] <ekr@jabber.org> An urgency value sounds totally different from a priority!
[13:11:52] chi.jiun.su joins the room
[13:14:32] <Martin Thomson> Why does webex not automatically mute when joining?
[13:15:00] <tfpauly> I assume that’s a setting that could be configured?
[13:15:05] <tfpauly> Can you see things now, Martin?
[13:15:23] <ekr@jabber.org> It is a setting. Mine was muted
[13:15:26] <Martin Thomson> It is, but I don't recall ever having that control (but I don't download software for this)
[13:15:42] <Martin Thomson> After rejoining, I was not muted
[13:15:55] <tfpauly> Huh, odd
[13:15:59] <mcr> there is a setting that the meeting owner can set that does this. Under Advanced settings.
[13:16:52] <mcr> uhm. There used to be.
[13:18:25] ChrisBox leaves the room
[13:26:58] <Martin Thomson> Not saying that there isn't value from learning from experiments, but we're uniquely weird.
[13:27:08] <Martin Thomson> Simpler than HTTP/2 is not the bar we should be aiming for.
[13:31:13] ChrisBox joins the room
[13:31:48] ChrisBox leaves the room
[13:32:49] ChrisBox joins the room
[13:33:57] <Martin Thomson> lpardue: this is unique to this: all other instances (I think) of cross-stream identification occur after the stream ID has been retrieved from the transport layer
[13:38:17] ChrisBox leaves the room
[13:39:27] <ekr@jabber.org> Why not just have an extension that indicates the priorities version?
[13:39:46] <Martin Thomson> Settings arrive too late for this to be useful early on.
[13:39:47] ChrisBox joins the room
[13:39:55] <ekr@jabber.org> TP :)
[13:40:13] <Martin Thomson> They are out at my local supermarket
[13:43:43] <ekr@jabber.org> Are others seeing a slide right now?
[13:44:01] <Martin Thomson> Let us spend time on something else and let the editors decide.
[13:44:03] <tfpauly> Yes?
[13:44:10] <Martin Thomson> Slide 16
[13:44:29] <ekr@jabber.org> I am just seeing whitespace
[13:44:45] <Martin Thomson> Webex does that
[13:45:43] <Magnus Westerlund> @ekr reloading the session is the fix that I know work.
[13:45:46] <ekr@jabber.org> Yeah, I did
[13:48:35] <mcr> we should have a round of introductions during the Administrivia, where everyone who is expected to speak says hello.
[13:48:59] <mcr> of course, things can break afterwards, but at least we'd find this kind of problem earlier.
[13:49:01] <Martin Thomson> https://github.com/httpwg/wg-materials/blob/gh-pages/interim-20-05/user_defined_resource.pdf
[13:55:10] <mcr> I am very interested in this draft (because I need it!). I heard many valid concerns about TLS1.3 abilities from EKR. I think that it could be that we want two solutions: a simple solution like this where we actually then consider restricting (subset) TLS1.3 abilities. Even without the reverse proxy, some of those TLS activities (like changing auth) might not have clear meanings. We then need a more complex document that makes it clear how the advanced uses of TLS1.3 can interact with HTTP, and once we figure that out, we'll know how to pass that activity from reverse-proxy to application.
[13:55:21] <mcr> [I have to chair a RATS call in five minutes]
[13:55:47] <Martin Thomson> mcr: what is a "TLS 1.3 ability"?
[13:56:02] <mcr> I'm prehaps using an incorrect term.
[13:56:08] <Martin Thomson> and why is 1.3 relevant here?
[13:56:08] <ekr@jabber.org> Actually, I think nearly all the issues are generic TLS issues
[13:56:17] <mcr> Things that TLS 1.3 can do that TLS 1.2 either doesn't, or infrequently does.
[13:56:19] <ekr@jabber.org> (resumption, renegotiation)
[13:56:32] <Martin Thomson> 1.3 can only do less than 1.2, for the most part
[13:56:51] <mcr> I may be overstating the 1.3-ness of the issue then.
[13:57:02] <ekr@jabber.org> Ironically, one might be able to use HTTP Request signing
[13:57:20] <ekr@jabber.org> to bind the middlebox's endorsement
[13:57:35] <Martin Thomson> ekr if you understood what it meant to sign the request
[13:57:36] <mcr> binding the middlebox, that's interesting.
[13:57:45] <ekr@jabber.org> @MT: right, of course
[13:57:45] <mcr> I trhought you meant the client could do it end to end.
[13:57:58] <Martin Thomson> mcr, yes that is what he was suggesting
[13:58:09] <Martin Thomson> exported authenticators work for this, but not e2e
[13:58:25] <ekr@jabber.org> @mcr: potentially a good idea for the future, but the idea here is to enable the existing clunky uses of TLS client auth
[13:58:56] <Martin Thomson> but exported authenticators are too heavyweight, a channel binding is fine, but the h2 certs draft has many of the pieces you might want
[13:59:17] <ekr@jabber.org> @MT: you mean a channel binding to the proxy->server connection?
[13:59:19] <mcr> yeah, so that's my point: I think we have an immediate need, and restricting some forms for renegotiation so that the client certificate can not change, might be appropriate.
[13:59:20] ChrisBox leaves the room
[13:59:50] <mcr> but, I think that we need to deal with the more complex back end issue.
[13:59:55] <mcr> okay, I gotta go....
[14:00:01] <Martin Thomson> ekr, yeah, if you have to ensure that the middlebox makes the assertion unambiguously, then having it bound to the connection somehow helps avoid the copy-paste thing
[14:00:08] <ekr@jabber.org> yeah
[14:00:08] felixh joins the room
[14:01:03] James Gruessing leaves the room
[14:01:10] <mcr> I need the certificate chain, but that's as fancy as I care about.
[14:01:30] <mcr> I also need ability to delegate cert chain validation to application via some reply answer.
[14:02:28] <ekr@jabber.org> To recap what I said on list: I'm happy to have a code point assignment for this that isn't standards track more or less as-is. But I'm uncomfortable with this being published as standards track in this form.
[14:03:22] <ekr@jabber.org> And so if we know that many people who want this mechanism are totally opposed to anything fancier, then we might as well just do the code point assignment now.
[14:03:59] ChrisBox joins the room
[14:04:11] ekr@jabber.org leaves the room
[14:08:29] <Martin Thomson> Was this draft discussed on the list at all?
[14:08:55] We had a couple of calls to discuss it, and a few responses.
[14:09:21] <Martin Thomson> OK, I finally found the thread (New Version Notification ..., very helpful)
[14:09:55] Ask the tools team to change the subject arrangement.
[14:11:47] ChrisBox leaves the room
[14:19:33] ChrisBox joins the room
[14:19:54] @cory - yes, but let's not create more problems :)
[14:20:00] ChrisBox leaves the room
[14:21:43] <Martin Thomson> Mark Nottingham: because Sydney is so much different than Melbourne...
[14:21:57] it used to be
[14:22:01] chi.jiun.su leaves the room
[14:22:06] Barry Leiba leaves the room
[14:22:20] felixh leaves the room
[14:22:48] Eliot gave me an amazing amount of detail about the history of Australian timezones as an aside.
[14:25:18] Mark Nottingham leaves the room
[14:25:35] <Martin Thomson> We should get him and dbaron in a room.
[14:25:42] Martin Thomson leaves the room
[14:38:35] lpardue leaves the room: Disconnected: BOSH client silent for over 60 seconds
[14:38:48] Magnus Westerlund leaves the room
[15:47:57] tfpauly leaves the room
[16:20:45] bdc leaves the room: Disconnected: closed