IETF
rtcweb@jabber.ietf.org
Friday, August 2, 2013< ^ >
gmaxwell has set the subject to: RTCWEB WG http://tools.ietf.org/wg/rtcweb/agenda | IETF 87 audio:  http://ietf87streaming.dnsalias.net/ietf/ietf871.m3u | IETF 87 slides: https://datatracker.ietf.org/meeting/87/materials.html#wg-rtcweb | Meetecho: http://www.meetecho.com/ietf87/rtcweb
Room Configuration
Room Occupants

GMT+0
[06:42:12] Wendy Seltzer leaves the room
[06:55:34] saghul joins the room
[07:01:50] stpeter joins the room
[07:08:20] Wendy Seltzer joins the room
[07:19:48] teco.boot joins the room
[07:38:09] saghul leaves the room
[08:34:02] tomkrist joins the room
[08:50:06] teco.boot joins the room
[09:09:15] teco.boot leaves the room
[09:11:11] tomkrist leaves the room
[09:11:19] tomkrist joins the room
[09:14:25] Lorenzo Miniero joins the room
[09:15:46] wolfgang.beck01@gmail.com joins the room
[09:16:20] Guest joins the room
[09:17:46] andrey.uzunov joins the room
[09:18:01] Jonathan Lennox joins the room
[09:18:24] gmandyam joins the room
[09:19:06] <Guest> Mic test ok
[09:19:18] Simon Romano joins the room
[09:19:38] jesup joins the room
[09:19:46] Guest has set the subject to: RTCWEB WG http://tools.ietf.org/wg/rtcweb/agenda | IETF 87 audio:  http://ietf87streaming.dnsalias.net/ietf/ietf874.m3u | IETF 87 slides: https://datatracker.ietf.org/meeting/87/materials.html#wg-rtcweb | Meetecho: http://www.meetecho.com/ietf87/rtcweb
[09:20:03] <Guest> Audio stream link was wrong
[09:20:04] <Lorenzo Miniero> for remotes, Meetecho available here http://www.meetecho.com/ietf87/rtcweb
[09:20:04] mreavy joins the room
[09:20:24] <jesup> meetecho doesn't seem to be up yet
[09:20:44] <Lorenzo Miniero> it should be up any minute now
[09:20:53] Xavier Marjou joins the room
[09:21:14] <Guest> IT should have been up 5 mins before the meeting
[09:21:22] stpeter joins the room
[09:21:32] <Lorenzo Miniero> it is up, we're updating the redirects
[09:21:46] Olle E. Johansson joins the room
[09:22:48] teco.boot joins the room
[09:22:49] gmandyam leaves the room
[09:22:52] <Guest> Up, but got this error: Unhandled event: rtcweb
[09:22:56] <jesup> Lorenzo Miniero: "Unhandled event tcweb"
[09:23:01] <jesup> rtcweb even
[09:23:13] tsuichi joins the room
[09:23:23] <Lorenzo Miniero> wring redirect I guess
[09:23:29] <Lorenzo Miniero> try this one instead: http://berlin.conf.meetecho.com/WebLite/login.jsp?ietf=rtcweb
[09:23:44] <Lorenzo Miniero> Slide 1: RTCWEB
[09:23:58] <Lorenzo Miniero> Current presenter: Chairs
[09:23:58] <Lorenzo Miniero> Slide 1: RTCWEB
[09:24:03] Roger Rabbit joins the room
[09:24:06] <Guest> That worked
[09:24:07] xavier marjou joins the room
[09:24:14] <Lorenzo Miniero> sorry about that!
[09:24:16] <Olle E. Johansson> We have a huge room and it's slowly filling up
[09:24:18] Maire Reavy joins the room
[09:24:20] <Lorenzo Miniero> the meeting hasn't started yet though
[09:24:20] Giri Mandyam joins the room
[09:24:26] <Lorenzo Miniero> so nothing missed :)
[09:24:27] Stephan Wenger joins the room
[09:24:34] <Lorenzo Miniero> starting now
[09:24:45] Antonin Maréchal joins the room
[09:25:05] Randell Jesup joins the room
[09:25:29] Antonin Maréchal leaves the room
[09:25:33] Antonin Maréchal joins the room
[09:25:40] <Lorenzo Miniero> Slide 2: Note Well
[09:25:43] pm joins the room
[09:25:47] Tobias joins the room
[09:25:48] Guillaume Leclanche joins the room
[09:25:54] Xavier Marjou leaves the room
[09:26:04] Maire Reavy leaves the room
[09:26:18] Tetsuya Hishiki joins the room
[09:26:25] Makkes joins the room
[09:26:28] Gavin Llewellyn joins the room
[09:26:39] <Lorenzo Miniero> Slide 4: Agenda Day 2
[09:26:52] Nora joins the room
[09:26:52] <Olle E. Johansson> Ted Hardie speaking
[09:26:54] Maire Reavy joins the room
[09:26:57] <Olle E. Johansson> as chair
[09:27:17] daniel@ngs.asipto.com joins the room
[09:27:18] <Olle E. Johansson> welcome to the rtcweb chat room! We have meetecho coverage of this session if you want slides and video
[09:27:31] <Guest> Great, now the audio is distorted. Isn't there a way to turn up the volume on the amplifier locally instead?
[09:27:37] <Lorenzo Miniero> so no "Slide X" typing for you today Olle :)
[09:27:46] <Olle E. Johansson> Horray!
[09:27:53] <Olle E. Johansson> Did we assign a jabber scribe?
[09:27:59] <Lorenzo Miniero> FYI, Meetecho available here: http://www.meetecho.com/ietf87/rtcweb (redirect fixed)
[09:28:11] <Guest> I heard Cullen forced someone.
[09:28:18] <Lorenzo Miniero> Guest I'm monitoring our audio feed and it doesn't sound distorted
[09:28:21] chris joins the room
[09:28:33] <Lorenzo Miniero> I think Bernard was asked to do notes
[09:28:44] <Lorenzo Miniero> Presentation stopped
[09:28:56] <Lorenzo Miniero> Slide 1: RTCWeb: Current Draft Status
[09:29:02] <Lorenzo Miniero> Current presenter: Cullen Jennings
[09:29:03] <Lorenzo Miniero> Slide 1: RTCWeb: Current Draft Status
[09:29:04] <Lorenzo Miniero> Slide 2: About the percent done numbers on follow
[09:30:00] <Lorenzo Miniero> Slide 3: Drafts 1 of 3
[09:31:43] <Olle E. Johansson> "What do we need to finish" - Cullen
[09:32:02] <Olle E. Johansson> Lines queueing up
[09:32:09] <Olle E. Johansson> Andy Hutton speaking
[09:32:26] <Olle E. Johansson> Firewall requirements document should be on the list - andy
[09:32:34] Stephan Wenger leaves the room
[09:33:45] <Olle E. Johansson> Tirumaleswar Konda talking
[09:34:43] <Olle E. Johansson> "That was really unpleasant" - Cullen
[09:34:46] <Jonathan Lennox> Microphone explodes.
[09:35:03] <Lorenzo Miniero> sabotage!
[09:35:20] Stephan Wenger joins the room
[09:35:25] <Olle E. Johansson> Martin Thomson talking
[09:36:06] <Olle E. Johansson> Harald Alvestrand in line
[09:36:57] <Olle E. Johansson> "I think the IETF procedures exist to get work done" - Harald
[09:37:20] <Olle E. Johansson> Gonzalo Camarillo talking
[09:37:34] <Olle E. Johansson> as AD
[09:37:56] <Olle E. Johansson> Bernard Aboba talking
[09:38:17] <Olle E. Johansson> note: If you want us to relay something to the room as a remote participant, please prefix your statement with "mic:"
[09:38:32] <Olle E. Johansson> Gonzalo talking again
[09:38:37] Spencer Dawkins joins the room
[09:38:44] Wendy Seltzer joins the room
[09:39:01] <Olle E. Johansson> Martin Thomson speaking
[09:39:34] <Olle E. Johansson> Dan Burnett talking
[09:40:21] <Olle E. Johansson> Procedure discussions :-)
[09:40:31] <Lorenzo Miniero> Slide 4: Drafts 2 of 3
[09:40:52] <Olle E. Johansson> Cullen talking
[09:40:59] <Olle E. Johansson> Martin Thomson speaking
[09:41:21] <Olle E. Johansson> Emil Ivov running to the mike
[09:42:19] <Lorenzo Miniero> Olle Johansson speaking
[09:42:38] gmaxwell joins the room
[09:42:48] <Olle E. Johansson> Andrew Allen talking
[09:42:52] <Guest> Cheers Olle!
[09:42:56] <Olle E. Johansson> Bernard Aboba talking
[09:43:02] <Olle E. Johansson> :-)
[09:43:11] <Lorenzo Miniero> scribes of the world, unite :)
[09:43:22] <Lorenzo Miniero> Slide 5: Drafts 3 of 3
[09:43:39] <Lorenzo Miniero> Slide 4: Drafts 2 of 3
[09:43:47] <Olle E. Johansson> Martin Thomson speaking
[09:44:21] <Olle E. Johansson> EKR getting close to the mic
[09:44:25] saghul joins the room
[09:44:25] <Lorenzo Miniero> Slide 5: Drafts 3 of 3
[09:44:38] <Olle E. Johansson> This is a long list of documents
[09:44:50] <Olle E. Johansson> Colin Perkins talking
[09:45:09] <Olle E. Johansson> "Please implement circuit breakers - if you have RTCP which you SHOULD!" - Colin Perkins
[09:45:32] <Olle E. Johansson> Jonathan Lennox coming soon
[09:45:36] <Olle E. Johansson> Now
[09:46:14] <Olle E. Johansson> "Are we going to do an updated version of this list that is even more depressing?" - Jonathan
[09:46:31] <Olle E. Johansson> Martin Thomson talking
[09:46:49] <Olle E. Johansson> app-id is missing, partial offer/answer is also missing
[09:47:09] <Olle E. Johansson> Jonathan Lennox again
[09:47:27] <Olle E. Johansson> Colin Perkins
[09:47:40] <Lorenzo Miniero> Slide 4: Drafts 2 of 3
[09:48:11] <Olle E. Johansson> Bernard speaking
[09:48:27] <Olle E. Johansson> Lorenzo: The slide says #5
[09:48:32] Yoshikazu GOTO joins the room
[09:48:42] <Lorenzo Miniero> Slide 5: Drafts 3 of 3
[09:48:49] <Lorenzo Miniero> missed the last trigger, sorry
[09:48:57] <Olle E. Johansson> Martin Thomson speaking
[09:50:17] <Olle E. Johansson> Btw, I'm scribing over NAT64 wifi network.
[09:50:44] <Olle E. Johansson> Next IETF, I expect the Scribes to use RTCweb data channels or something more sexy :-)
[09:51:02] <Olle E. Johansson> EKR saying yes yes yes by waving a hand
[09:51:13] <Lorenzo Miniero> Presentation stopped
[09:51:19] <Olle E. Johansson> Christer Holmberg coming up
[09:51:24] <Lorenzo Miniero> Slide 1: USE-CASES
[09:51:32] <Lorenzo Miniero> Current presenter: Christer Holmberg
[09:51:33] <Lorenzo Miniero> Slide 1: USE-CASES
[09:51:33] <Olle E. Johansson> With this big room, we have tons of empty chairs, but we're still quite a few people
[09:51:50] <Lorenzo Miniero> Slide 2: (1) CHANGES FROM -10
[09:51:53] <Lorenzo Miniero> Slide 3: (2) CHANGES FROM -10
[09:51:57] <Lorenzo Miniero> Slide 4: (3) NEXT STEP
[09:52:00] <Lorenzo Miniero> Slide 3: (2) CHANGES FROM -10
[09:52:03] <Lorenzo Miniero> Slide 4: (3) NEXT STEP
[09:52:13] <Lorenzo Miniero> "WGLC" :)
[09:53:12] <Olle E. Johansson> EKR speaking
[09:53:17] <Olle E. Johansson> The slide data makes the slides deprecated
[09:53:17] <Lorenzo Miniero> Slide 1: USE-CASES
[09:53:28] <Olle E. Johansson> date
[09:53:32] <Lorenzo Miniero> but the first slide is from the future
[09:53:39] <Lorenzo Miniero> Slide 4: (3) NEXT STEP
[09:53:45] <Olle E. Johansson> Amazing presenation
[09:53:47] <Guest> 2014 is far closer than 20013
[09:54:13] <Lorenzo Miniero> Presentation stopped
[09:54:19] <Olle E. Johansson> Data Channels coming up!!!
[09:54:24] <Lorenzo Miniero> Slide 1: Data    
 Channels    
 
[09:54:33] <Olle E. Johansson> Time for all the fun. Let's route IPv4 over rtcweb routers with data channels
[09:54:33] <Lorenzo Miniero> Current presenter: Michael Tuxen
[09:54:33] <Lorenzo Miniero> Slide 1: Data    
 Channels    
 
[09:54:46] <Lorenzo Miniero> Slide 2: Outline    
 
[09:54:59] <Olle E. Johansson> The page number is actually #1 on slide 2
[09:55:06] <Olle E. Johansson> Note to anyone following without meetecho
[09:55:14] <Lorenzo Miniero> Slide 3: Changes    
  draF-­‐ieG-­‐rtcweb-�
[09:55:15] <Olle E. Johansson> (You should be using meetecho :-) )
[09:55:23] Cullen Jennings joins the room
[09:55:33] <Lorenzo Miniero> yes you should :)
[09:55:37] <Lorenzo Miniero> good point anyway
[09:55:54] <Lorenzo Miniero> I guess the title slide is not in the numbering
[09:56:02] <Lorenzo Miniero> Slide 4: Changes    
  draF-­‐ieG-­‐rtcweb-�
[09:56:12] <Lorenzo Miniero> draft-ietf-rtcweb-data-protocol-00
[09:56:29] alfredh joins the room
[09:56:39] <Olle E. Johansson> Repeat: If you want anything relayed to the meeting from the jabber chat, please prefix your statement here with "mic:" - Thank you.
[09:57:05] <Lorenzo Miniero> Slide 5: Open    
 Issue    
 No    
 1:    
  IniRal
[09:57:11] <Lorenzo Miniero> Initial PMTU
[09:57:53] <Olle E. Johansson> EKR speaking (Eric Rescorla)
[09:58:42] <Olle E. Johansson> Randall Stewart speaking
[09:59:06] <Olle E. Johansson> "Eiiii yak" - Randall
[09:59:15] <Olle E. Johansson> Hadriel Kaplan speaking
[09:59:29] <Olle E. Johansson> Scribe note: It took a whlie for Hadriel to show up
[09:59:47] stpeter leaves the room: Offline
[10:01:07] <Olle E. Johansson> Justin Uberti talking
[10:01:20] <Olle E. Johansson> We're tunneling and tunneling and tunneling
[10:02:08] Neil Stratford joins the room
[10:02:55] Stephan Wenger leaves the room
[10:03:11] <Olle E. Johansson> Ekr talking
[10:03:37] <Olle E. Johansson> We have EKR, Martin Thomson, Justin Uberti and Hadriel hovering around the mike
[10:03:49] <Olle E. Johansson> Justin talking soon
[10:03:50] Neil Stratford leaves the room
[10:03:53] tamir joins the room
[10:04:07] <Olle E. Johansson> SCTP is running in user space - justin
[10:04:42] <jesup> mic: size of the SCTP layer should be 1280 minus worst-case headers (not actual headers)
[10:04:58] <Olle E. Johansson> ok
[10:05:27] <Guest> mic: I was under the impression that the lower layer stack is going to take care of packetization - meaning that you can send any number of bytes in the payload. So why does a browser need to know and bother with the low layers and do the packetization itself?
[10:05:33] <jesup> Olle E. Johansson: which is basically what Martin said
[10:06:09] Neil Stratford joins the room
[10:06:14] martin.thomson joins the room
[10:06:17] <Olle E. Johansson> done
[10:06:20] <Olle E. Johansson> hadriel
[10:06:26] <Olle E. Johansson> justin uberti speaking
[10:06:39] <Olle E. Johansson> hadriel kaplan speaking
[10:06:43] <martin.thomson> Can "Guest" please identify him/her/...self?
[10:07:16] <Lorenzo Miniero> capturing Michael with the webcam is becoming increasingly difficult! he's like a lion in the cage :)
[10:07:36] <Olle E. Johansson> Cullen talking as individual (in the line)
[10:08:20] andrey.uzunov leaves the room
[10:08:51] alfredh leaves the room
[10:09:01] <Olle E. Johansson> V4 networks with vpn activity got problems with 1280 mtu on ip layer - Cullen
[10:09:07] <Olle E. Johansson> 1200 is more reasonable - Cullen
[10:09:17] andrey.uzunov joins the room
[10:09:25] <gmaxwell> Yo dawg, I heard you like tunnels.
[10:09:27] <Olle E. Johansson> Jonathan Lennox talking
[10:09:29] <jesup> I'm good with 1200
[10:10:36] <Olle E. Johansson> Ted making a summary
[10:10:37] <Lorenzo Miniero> Slide 6: Open    
 Issue    
 No    
 2:    
  Data
[10:10:47] <Lorenzo Miniero> Data Channel Priority
[10:11:18] Cullen Jennings leaves the room
[10:11:46] martin.thomson leaves the room
[10:11:54] <Olle E. Johansson> Richard Ejzak speaking
[10:12:17] alfredh joins the room
[10:12:27] <Olle E. Johansson> IETF staff has tie-dye friday. Richard and I have hawaiian shirt friday
[10:12:36] <Olle E. Johansson> Martin thomson next
[10:12:53] <Lorenzo Miniero> I'll buy one in Honolulu!
[10:13:29] <Guest> mic: I'd suggest dropping 'priority' as a whole - I saw no practical use as it's unlikely that two data channels would be opened for the same 'stuff' (and hence there's no need for priority).
[10:13:30] <Olle E. Johansson> Randall Stewart next
[10:13:55] <Olle E. Johansson> bernard
[10:14:27] <Lorenzo Miniero> Guest: what's your name, for the records?
[10:14:41] <Olle E. Johansson> martin thomson disagree
[10:14:49] Tim Panton joins the room
[10:15:06] <Olle E. Johansson> It is important for the Note Well to have some sort of name
[10:15:56] sal joins the room
[10:16:14] <Olle E. Johansson> Ted Hardie at the mic as individual
[10:16:26] tobia joins the room
[10:16:35] <Olle E. Johansson> Dan Druta in line after Ted
[10:16:57] <Olle E. Johansson> Guest: relayed your statement
[10:17:00] Cullen Jennings joins the room
[10:17:34] <Guest> +1 to what Ted is saying; apps know exactly what they are exchanging and how important that exchange is.
[10:17:39] <Olle E. Johansson> We will have to discuss how anonymous participation works with the note well and legal issues of the IETF. Please don't take that personal, it's just a matter of procedure.
[10:18:01] <Olle E. Johansson> Mike line growing
[10:19:16] <Olle E. Johansson> Ted relays Jonathan in the room
[10:19:19] <Olle E. Johansson> Dan Druta
[10:19:30] <Guest> Not taking it personal; I do not see a conflict with anonymous participation and note well however.
[10:19:39] <xavier marjou> Dear Guest, I anticipate that in the future, my loggin procedure will take more time in order to avoid anonymous people like you :-(
[10:19:46] <Guest> If there's one, please let me know and I'll drop off.
[10:19:49] teco.boot leaves the room
[10:20:24] <Olle E. Johansson> Guest: As for now, we're fine with you participating and have understanding. We just realize that we will have to handle this situation somehow.
[10:20:35] <Olle E. Johansson> Martin Thomson at mike
[10:22:13] <Olle E. Johansson> Ted discussing from chair desk
[10:24:05] stpeter joins the room
[10:24:05] <Olle E. Johansson> Richard Ejzak
[10:24:58] hta joins the room
[10:25:11] <Olle E. Johansson> Cullen speaking as Dr Jennings and esteemed friend of the chair, but not as chair
[10:25:22] <Guest> lo
[10:25:24] <Guest> l
[10:25:32] <Olle E. Johansson> Jonathan Lennox
[10:26:20] <Olle E. Johansson> Michael Welzl speaking
[10:26:44] <Olle E. Johansson> Congestion control is handled by rmcat wg, and Michael is behind one of the draft proposals
[10:27:34] Nora leaves the room
[10:27:35] <Olle E. Johansson> Martin THomson
[10:27:50] <Olle E. Johansson> "I did not say hum people. Please do not hum!" - Ted
[10:28:09] <Guest> How much does "the most expensive bottle of champagne in Berlin' cost?
[10:28:26] <Olle E. Johansson> That's something to find out :-)
[10:28:26] <Jonathan Lennox> It's pretty clear it's not going to be an issue.
[10:28:28] <Guest> *hum* for underspecified, still proposing removing it.
[10:28:32] <Lorenzo Miniero> there's no such thing in Berlin, just beer ;)
[10:28:41] <jesup> I agree with Ted, but also with Martin - we could say SHOULD use a fair scheduling such weighted fair queuing, and SHOULD NOT use strict priorities.
[10:29:03] <jesup> hum (in favor)
[10:29:24] <Olle E. Johansson> Martin Thomson speaking trying to explain what Ted thought.
[10:29:50] <Olle E. Johansson> Ted speaking as chair
[10:30:06] stpeter leaves the room: Offline
[10:30:10] <Lorenzo Miniero> Slide 7: Open    
 Issue    
 No    
 3:    
  Handli
[10:30:18] <Lorenzo Miniero> Handling of Early Data
[10:30:46] <Olle E. Johansson> 10 seconds is an eternity said the SIP guy.
[10:32:18] <Olle E. Johansson> lines are building at the mics
[10:32:26] <Olle E. Johansson> Hadriel first, Jonathan next
[10:32:35] <Olle E. Johansson> Bernard after jonathan
[10:32:45] <Olle E. Johansson> Oh, we have queue reordering
[10:32:59] <Olle E. Johansson> Jonathan lennox
[10:33:02] Tobias leaves the room: Disconnected: Received SIGTERM
[10:33:06] <jesup> Olle E. Johansson: weighted fair queuing?
[10:33:06] Tobias joins the room
[10:33:10] <Olle E. Johansson> Hadriel Kaplan
[10:33:11] <Guest> mic: Wouldn't this open the door for DoS attacks? I'd require dropping everything until the DATA_CHANNEL_OPEN is received. I guess unordered will be used only in case reliable transmission is not mission-critical.
[10:33:45] <Olle E. Johansson> in queue
[10:33:50] <Guest> Cheers.
[10:34:26] tobia leaves the room
[10:34:44] andrey.uzunov leaves the room
[10:34:47] Stephan Wenger joins the room
[10:35:10] <Olle E. Johansson> hadriel waves arms, can't relay that
[10:35:24] <Olle E. Johansson> ted speaking from chair desk
[10:36:48] <Olle E. Johansson> discussion between speaker, Ted as chair and Hadriel Kaplan
[10:37:05] <Olle E. Johansson> Guest: Still in queue
[10:37:09] <Olle E. Johansson> Jonathan speaking
[10:37:30] Stephan Wenger leaves the room
[10:37:34] Stephan Wenger joins the room
[10:37:35] <Olle E. Johansson> Queue is growing
[10:38:12] alfredh joins the room
[10:38:20] <Olle E. Johansson> "If it's unreliable you just drop it" - Jonathan
[10:38:35] <Guest> +1 to Jonathan
[10:38:37] <jesup> mic: if minimum buffer is not specified and it's unordered, then the application must either a) support loss of packets at the opening of the channel, even if it's a reliable channel, or b) the application must implement a handshake on it's own
[10:38:38] <Olle E. Johansson> Martin Thomson at mic
[10:38:44] <Olle E. Johansson> Ok, jesup
[10:39:17] <Olle E. Johansson> Bernard Aboba next in line at mic
[10:39:39] <Olle E. Johansson> Martin speaking still
[10:40:58] <Olle E. Johansson> Bernard speaking now
[10:42:08] wolfgang.beck01@gmail.com leaves the room
[10:42:23] wolfgang.beck01@gmail.com joins the room
[10:42:46] wolfgang.beck01@gmail.com leaves the room
[10:42:49] <jesup> mic: make that my proposal
[10:43:18] martin.thomson joins the room
[10:43:34] <Guest> To get the TBD out of the document, you can say: Endpoints SHOULD expect their packets be dropped until the destination receives the DATA_CHANNEL_OPEN; the destination MAY perform a friendly buffering, however the endpoints MUST NOT rely on that.
[10:43:52] <Guest> @Cullen: Fair enough, understood.
[10:44:01] <Olle E. Johansson> done
[10:44:13] <Olle E. Johansson> Greg Maxwell speaking
[10:44:19] <jesup> thanks
[10:44:22] <Olle E. Johansson> :-)
[10:44:35] <Olle E. Johansson> @jesup: You got positive feedback in the room
[10:44:35] Stephan Wenger leaves the room
[10:44:41] <Olle E. Johansson> Justin Uberti
[10:45:44] <jesup> To be clear, I was proposing no minimum specified, and thus applications must (if unordered)  do my a or b
[10:45:59] <martin.thomson> I believe that the proposal I would make is: a) buffer as much as you can/like up to whatever limits you feel appropriate to the constraints of your application, and b) an overflow should be considered as an error in the stream and should generate an error.
[10:46:16] Giri Mandyam leaves the room
[10:46:26] Marcon Jérôme joins the room
[10:46:26] <Olle E. Johansson> Jonathan Lennox speaking
[10:46:26] <Guest> Regarding what Ted just said: if you have to make sure that the channel is open, you better use an ordered channel.
[10:46:43] burn joins the room
[10:47:03] <Olle E. Johansson> Admin note: Don't forget to prefix with "mic:" if you want me to relay
[10:47:22] <Olle E. Johansson> Eric Rescorla speaking
[10:47:29] gmandyam joins the room
[10:47:33] <Olle E. Johansson> "This is sort of making me a sad panda" - EKR
[10:47:43] sal leaves the room
[10:47:52] alfredh leaves the room
[10:48:05] <martin.thomson> To reiterate: unreliable packets are just dropped, so this is never an issue, and ordered packets are protected by flow control windows so there is control over the resource commitment.  Thus, reliable + unordered is the only case where this causes problems.
Furthermore, if there is a mismatch between peers about whether this is in-band/out-of-band negotiation, you are just permanently screwed.
[10:48:25] <Olle E. Johansson> Martin: I am not going to relay you, sorry. ;-)
[10:48:33] <hta> what I don't understand, this may be an API question.... why does the app layer have to take the handover of the unordered buffers?
[10:48:47] <Guest> Can we forbid the reliable + unordered combination?
[10:48:57] <Guest> Makes little sense.
[10:49:17] wolfgang.beck01@gmail.com joins the room
[10:49:21] <Olle E. Johansson> Cullen Jennings next on mic
[10:49:25] <martin.thomson> It makes some sense.
[10:49:25] <Guest> (no need to relay that)
[10:49:39] <Guest> All right, ignore what I said then
[10:49:39] Giri Mandyam joins the room
[10:49:46] <Olle E. Johansson> Cullen Jennings as individual
[10:49:52] <martin.thomson> This is one of the use cases that drove the development of TCP minion APIs.
[10:50:59] martin.thomson leaves the room
[10:51:37] <Guest> mic: Regarding Cullen's proposal: ok, but and endpoint should not be required to wait for that response (it could if it wanted to).
[10:52:07] <Olle E. Johansson> ok, queued
[10:52:25] <Olle E. Johansson> hadriel kaplan talking loudly
[10:53:22] <Guest> mic: Tad, data channel reset is used to close the channel. How would that be interpreted on the other end that sent an OPEN message?
[10:53:32] <Guest> Ted, sorry for the typo
[10:53:35] <gmaxwell> What this really wants in partial ordered delivery.  But IIRC there is no way to really achieve that here. ("lets go digging up all the protocol layers!")
[10:54:01] <Olle E. Johansson> Harald Alvestrand next on mic
[10:54:45] <Olle E. Johansson> "I was about to give another proposal but like this one better. I like it. "  - harald
[10:54:48] <Olle E. Johansson> Martin Thomson
[10:56:08] <hta> There might be an argument for having a minimum buffer of 1 packet. Would be interesting to run some statistics on loss rate vs channel failure rate for various buffer sizes.
[10:56:10] <jesup> mic: we had planned to propose the reset (next slide?)  It does mean that any application that uses unordered MUST be ready to accept a open-time-reset-and-retry at the application level, or use a protocol where the side receiving the OPEN handshakes
[10:56:31] <gmaxwell> this is starting to sound like implementing your own reliable delivery in your application. :(
[10:56:48] martin.thomson joins the room
[10:56:49] <Olle E. Johansson> relayed
[10:56:54] <Olle E. Johansson> speaker answering
[10:56:57] <jesup> gmaxwell: only of the open handshake
[10:56:57] <hta> gmaxwell: just means that you have to deal with failure. That's necessary with TCP too (TCP RST).
[10:57:06] <Olle E. Johansson> Erik Rescorla speaking
[10:57:18] <jesup> ollie: I send a mic: while you wee up
[10:57:21] <jesup> were
[10:57:30] <Olle E. Johansson> Sorry missed that. Back in line.
[10:57:31] <Guest> Cheers, Olle!
[10:57:31] John Leslie joins the room
[10:58:04] <martin.thomson> A suggestion, in addition to the design Ted has described, is to have a client implementation re-attempt to create a channel if this error occurs.
[10:58:05] <gmaxwell> hta: I realize this, ... it was one of the points I was trying to make when I spoke. "if the network sucks enough you will fail".  But this is sounding like something everyone will need to deal with which isn't so great when you have a protocol that does reliable delivery already.
[10:58:28] <gmaxwell> this also means that some dumb applications may hammer.
[10:58:52] <Olle E. Johansson> Line building up
[10:58:55] <martin.thomson> This avoids the high probability of failure associated with packet loss.
[10:59:13] Cullen Jennings leaves the room
[10:59:49] <Olle E. Johansson> Ted and EKR discussing
[10:59:59] <Olle E. Johansson> Speaker talking
[11:00:12] <Olle E. Johansson> @jesup: I'm stil in line
[11:01:07] Tobias leaves the room
[11:01:26] <Guest> mic: Ted's proposal is flawed in case the buffer size is low (or 0) - I would be required to flood the stream with resets every time a new packet comes in.
[11:02:32] <martin.thomson> Not strictly, reset is only required once.
[11:02:33] <gmaxwell> If the buffering is something like 2x rtt you'll ~usually survive the case where the open is dropped and retransmitted.
[11:02:38] Cullen Jennings joins the room
[11:02:52] <Lorenzo Miniero> Slide 8: Open    
 Issue    
 No    
 4:    
  Conges
[11:03:13] <Guest> @martin: how'd you know that without having received the OPEN message?
[11:03:37] <Olle E. Johansson> @jesup and @guest done!
[11:03:42] <Guest> Thanks!
[11:03:44] <jesup> thanks
[11:03:49] <Olle E. Johansson> Bernard talked
[11:03:53] <Olle E. Johansson> hadriel Kaplan talking
[11:03:56] <Olle E. Johansson> Long line
[11:03:57] <gmaxwell> and thats a rather small amount of memory. e.g. for a gigabit link and a 200ms RTT you're talking about say 62megabytes regardless of the number of concurrent connections.
[11:04:00] <martin.thomson> Another thing is the fact that the a likely scenario is one where the first packets on the channel/stream are in the same IP packet as the open.
[11:04:12] <Olle E. Johansson> Paul Kyzviat coming up after Hadriel
[11:04:47] <hta> isn't there a channel open message at the SCTP protocol level, too? So if you reset the channel, the SCTP machinery knows that you don't want it any more?
[11:05:00] <Olle E. Johansson> Screaming in the line
[11:05:21] <Olle E. Johansson> Unordered and unreliable comments
[11:05:27] <jesup> hta: no stream open in SCTP
[11:05:29] <martin.thomson> @Guest: Interesting point.  That's a good argument for requiring at least a small amount of memory for this.
[11:06:10] <gmaxwell> lets just revise SCTP to allow for partial ordering. :P
[11:06:15] <Olle E. Johansson> 25 minutes to the end of the IETF #87 but not the end of the world.
[11:06:19] <Lorenzo Miniero> michael is getting closer and closer to the mic line!
[11:06:35] <Guest> I like Cullen's proposal better: send an OPEN confirmation, but do not require waiting for that.
[11:06:35] <Simon Romano> :-)
[11:06:36] <Olle E. Johansson> We need to fence him off :-)
[11:07:04] <Lorenzo Miniero> ;)
[11:07:17] <jesup> martin/guest: I think data that arrives on a channel that was reset could be dropped silently.  Resets are handshaked at the SCTP level
[11:07:19] <gmaxwell> "it's possible to fix"  well heck, lets just remove all the reliable stuff entirely, it's possible for applications to handle that themselves.
[11:07:21] <Olle E. Johansson> Paul Kyzivat speaking
[11:08:12] <martin.thomson> @jesup: good point, and add to that the fact that the flow control potentially comes into play...
[11:08:25] <hta> what's the sound of one hand shaking....?
[11:08:27] <Olle E. Johansson> Randall Stewart speaking
[11:08:34] <burn> isn't a one-way handshake a slap?
[11:09:06] <Olle E. Johansson> What is a three way then, burn? Shake and slap?
[11:09:23] <Olle E. Johansson> Jonathan Lennox speaking
[11:09:30] <burn> @olle:  a fist-fight, I imagine
[11:09:33] <jesup> mic: Proposal: drop unordered reliable  :-)  
[11:09:55] <Guest> mic: +1 to drop unordered reliable
[11:09:58] <Olle E. Johansson> jesup: Ack - I'm virtually in the queue
[11:09:59] gmaxwell stabs
[11:10:01] <Olle E. Johansson> Guest: ACK
[11:10:05] <jesup> mic: that's not *entirely* fascitious
[11:10:08] <martin.thomson> i'll realy that
[11:10:16] gmaxwell stabs not entirely fascitiously
[11:10:20] <martin.thomson> wow, that was bad: I'll relay that
[11:10:23] alfredh leaves the room
[11:10:44] <Olle E. Johansson> i am not sure how to pronounce "fascitious" in Swenglish
[11:11:01] Spencer Dawkins leaves the room
[11:11:36] <Olle E. Johansson> EKR in line. Drop baudrate, replay in slow-mo and you will get it :-)
[11:11:55] <Olle E. Johansson> Oh, he is slow now. Fantastic.
[11:12:06] <Olle E. Johansson> Speed going up. Hang on.
[11:12:07] <Lorenzo Miniero> yes, but accellerating :)
[11:12:11] <gmaxwell> slowstart.
[11:12:26] <gmaxwell> stop acking! window full!
[11:12:27] <burn> it's less about speed than dropped packets
[11:12:44] <Olle E. Johansson> Jerome Marcon next in line
[11:12:56] <Guest> Do words/min measurements exist? ;-)
[11:13:18] <gmaxwell> what is the problem with ordering until something comes back?
[11:13:24] <Olle E. Johansson> Jerome speaking
[11:13:33] <Jonathan Lennox> gmaxwell: defining what the "something" is, I think.
[11:13:53] <gmaxwell> Jonathan Lennox: the open is reliably delivered, it will be acked.
[11:13:57] <gmaxwell> Thats something.
[11:13:58] <Cullen Jennings> @greg, as soon as you say ordered, it means there can be a bunch more latency until you get into unorded mode
[11:13:58] <Makkes> gmaxwell: because this is a horrible API
[11:14:19] <Guest> It sounds as if the one-way handshake is set in stone.
[11:14:27] <gmaxwell> Cullen Jennings: can be in the case where the 0 buffer case would _fail_ and need to be resignaled by the app.
[11:14:45] <Makkes> Guest: the agreement on one-way handshake is shaking, I believe
[11:14:56] <gmaxwell> Guest: needing a two way handshake before transmission would stink for latency.
[11:15:46] <Jonathan Lennox> gmaxwell: it's acked in the SCTP stack, but that ack isn't normally exposed to the SCTP stack's API.   That was my suggestion, but it requires SCTP stack changes.
[11:15:48] <Olle E. Johansson> Everyone smiled
[11:15:49] EKR joins the room
[11:16:05] <Guest> Thanks!
[11:16:11] <Olle E. Johansson> Ted speaking as a chair
[11:16:43] <jesup> hum agree
[11:16:50] <mreavy> hum (agree)
[11:16:56] Giri Mandyam leaves the room
[11:17:01] <Olle E. Johansson> THanks for the humming!
[11:17:13] <Lorenzo Miniero> Slide 8: Open    
 Issue    
 No    
 4:    
  Conges
[11:17:18] <Lorenzo Miniero> Congestion Control
[11:17:23] <Olle E. Johansson> Was the humming unordered and reliable? - Miconda in the room
[11:17:46] <Olle E. Johansson> Speaker got back in control and we have 14 minutes left.
[11:19:13] <Olle E. Johansson> MIchael describing Congestion Control
[11:19:36] <Lorenzo Miniero> only two others slides left
[11:19:40] <Olle E. Johansson> — powering up all my devices for the flight home ---
[11:19:47] <Lorenzo Miniero> (excluding this one)
[11:20:00] <Olle E. Johansson> With the number of minutes per slide that's stressful
[11:20:04] <Olle E. Johansson> Michael speaking
[11:20:19] <Olle E. Johansson> Michael Welzl
[11:20:23] RjS joins the room
[11:21:13] <Olle E. Johansson> Martin Thomson speaking at mic
[11:21:24] <Olle E. Johansson> "I think this is a sensible plan" - Martin
[11:21:37] burn leaves the room
[11:21:39] <Olle E. Johansson> Bernard Aboba with a question
[11:21:55] <Olle E. Johansson> "That seems sucky" - Bernard
[11:21:59] tamir leaves the room
[11:22:09] <Olle E. Johansson> Jesup - you were mentioned by Michael here
[11:23:25] <Guest> mic: Can we add an initial "congestion window size" to the DATA_CHANNEL_OPEN message (and the ACK)?
[11:23:35] Olle E. Johansson leaves the room: Replaced by new connection
[11:23:35] Olle E. Johansson joins the room
[11:23:48] <Olle E. Johansson> Colin perkins speaking
[11:23:54] Olle E. Johansson is now known as oej (scribe)
[11:24:29] <oej (scribe)> Zahed Sarker speaking
[11:25:14] <oej (scribe)> Ted calls a hum
[11:25:40] <Lorenzo Miniero> Slide 9: Open    
 Issue    
 No    
 5:    
  Fragme
[11:25:45] <oej (scribe)> Moving to issue #5
[11:26:02] <oej (scribe)> Michael taking control of the floor again
[11:26:09] <Guest> We moved on; scratch that mic request.
[11:26:22] <oej (scribe)> I think Martin relayed it.
[11:27:00] <oej (scribe)> We are making progress here - at least I feel that way.
[11:27:07] <oej (scribe)> Cullen as individual lining up at the mic
[11:27:35] <oej (scribe)> Randall Stewart coming up next
[11:28:37] wolfgang.beck01@gmail.com leaves the room
[11:28:37] <oej (scribe)> Ted as chair asking questions
[11:29:13] <oej (scribe)> Just a few minutes left.
[11:29:19] <Lorenzo Miniero> Slide 10: Open    
 Issues    
 No    
 6:    
  Messa
[11:29:21] <oej (scribe)> (with the bonus allocation from the chairs)
[11:29:25] <Lorenzo Miniero> Message Size Limitations
[11:30:02] <Lorenzo Miniero> Slide 9: Open    
 Issue    
 No    
 5:    
  Fragme
[11:30:13] <oej (scribe)> We are functionally out of time.
[11:30:28] <oej (scribe)> Ted stretches the time (I warned you that they had extra allocation!)
[11:30:52] <martin.thomson> If you want websocket-compatibility, you NEED an 8 exabyte limit (I think that is what 2^63 translates into).
[11:31:05] <oej (scribe)> Martin Thomson
[11:31:07] <Lorenzo Miniero> not too much I hope: I need to get on a plane and parachute directly on a beach!
[11:31:33] <oej (scribe)> @lorenzo without a proper Hawaiian shirt? That's daring!
[11:31:46] <oej (scribe)> Ted calls a hum
[11:32:09] <jesup> hum willing to live with limits for now
[11:32:23] Wendy Seltzer leaves the room
[11:32:24] <oej (scribe)> Chairs disagreeing
[11:32:38] Wendy Seltzer joins the room
[11:32:51] EKR leaves the room
[11:32:52] <oej (scribe)> First hum:
[11:33:03] <jesup> hum willing to live with limits
[11:33:32] <Guest> *humming for first* (willing to live with limit)
[11:33:39] <mreavy> hum for first option
[11:34:09] martin.thomson leaves the room
[11:34:09] <oej (scribe)> thanks to everyone
[11:34:15] Makkes leaves the room: Machine going to sleep
[11:34:16] saghul leaves the room
[11:34:16] <oej (scribe)> We are closing the session now.
[11:34:20] RjS leaves the room
[11:34:24] chris leaves the room
[11:34:27] <Guest> Thank you all!
[11:34:29] <oej (scribe)> See you on the list soon!
[11:34:32] <Lorenzo Miniero> done!
[11:34:38] <Lorenzo Miniero> we'll make the recordings available soon
[11:34:40] xavier marjou leaves the room
[11:34:41] <Lorenzo Miniero> see you in Vancouver!
[11:34:44] <Lorenzo Miniero> Presentation stopped
[11:34:46] <oej (scribe)> I want parachute pictures Lorenzo!
[11:34:51] Roger Rabbit leaves the room
[11:34:52] Yoshikazu GOTO leaves the room
[11:34:53] Antonin Maréchal leaves the room
[11:34:53] gmandyam leaves the room
[11:34:53] tsuichi leaves the room
[11:34:56] Guest leaves the room
[11:35:07] Simon Romano leaves the room
[11:35:13] <Lorenzo Miniero> without an hawaiian t.shirt? you're right, I'd be crazy!
[11:35:30] Neil Stratford leaves the room
[11:35:36] Maire Reavy leaves the room
[11:35:37] Tim Panton leaves the room
[11:35:38] Randell Jesup leaves the room
[11:35:39] Tetsuya Hishiki leaves the room
[11:35:43] Gavin Llewellyn leaves the room
[11:35:46] pm leaves the room
[11:35:46] oej (scribe) leaves the room
[11:35:52] John Leslie leaves the room
[11:36:01] daniel@ngs.asipto.com leaves the room: Disconnected: session closed
[11:36:02] tomkrist leaves the room
[11:36:15] Guillaume Leclanche leaves the room
[11:36:33] Lorenzo Miniero leaves the room
[11:38:16] Cullen Jennings leaves the room
[11:40:23] Wendy Seltzer leaves the room
[11:41:08] hta leaves the room
[11:43:16] hta joins the room
[11:47:54] RjS joins the room
[11:51:05] gmaxwell leaves the room
[12:00:55] Jonathan Lennox leaves the room
[12:01:07] hta leaves the room
[12:01:13] RjS leaves the room
[12:01:34] Marcon Jérôme leaves the room
[12:04:02] hta joins the room
[12:05:05] Marcon Jérôme joins the room
[12:05:28] hta leaves the room
[12:23:47] Jonathan Lennox joins the room
[12:29:26] Jonathan Lennox leaves the room
[13:04:09] Marcon Jérôme leaves the room
[13:04:26] jesup joins the room
[13:08:21] Jonathan Lennox joins the room
[13:09:30] jesup leaves the room
[13:11:19] daniel@ngs.asipto.com joins the room
[13:33:27] Jonathan Lennox leaves the room
[13:33:59] Jonathan Lennox joins the room
[13:46:39] Wendy Seltzer joins the room
[13:48:57] Wendy Seltzer leaves the room
[13:49:06] jesup joins the room
[13:57:15] Wendy Seltzer joins the room
[13:58:30] jesup leaves the room
[13:59:29] Wendy Seltzer leaves the room
[13:59:49] Wendy Seltzer joins the room
[14:01:59] Wendy Seltzer leaves the room
[14:02:59] Jonathan Lennox leaves the room
[14:30:41] EKR joins the room
[14:32:54] daniel@ngs.asipto.com leaves the room: Disconnected: session closed
[14:46:43] EKR leaves the room
[14:51:32] daniel@ngs.asipto.com joins the room
[14:51:44] daniel@ngs.asipto.com leaves the room: Disconnected: Replaced by new connection
[14:51:44] daniel@ngs.asipto.com joins the room
[14:56:37] EKR joins the room
[14:58:58] EKR leaves the room
[15:11:42] daniel@ngs.asipto.com leaves the room: Disconnected: session closed
[17:21:50] Tobias joins the room
[18:12:16] Tobias leaves the room
[19:10:39] jesup leaves the room
[19:53:15] Jonathan Lennox joins the room
[19:54:25] Jonathan Lennox leaves the room
[20:45:24] daniel@ngs.asipto.com joins the room
[20:59:28] daniel@ngs.asipto.com leaves the room
[21:36:23] Wendy Seltzer joins the room
Powered by ejabberd Powered by Erlang Valid XHTML 1.0 Transitional Valid CSS!