[03:12:55] --- ian.mcdonald has joined
[03:13:12] <ian.mcdonald> testing that room is working
[03:13:46] --- ian.mcdonald has left
[03:14:14] --- ian.mcdonald has joined
[03:14:36] --- ian.mcdonald has left
[09:51:04] --- gorryf has joined
[10:10:35] --- gorryf has left
[12:01:23] --- lars has joined
[12:01:29] <lars> yep
[12:45:10] --- arjuna_sai has joined
[12:49:47] --- arjuna_sai has left
[12:53:30] --- arjuna_sai has joined
[12:55:05] --- ian.mcdonald has joined
[12:56:44] --- csp has joined
[13:00:49] --- monden has joined
[13:02:04] --- monden has left
[13:03:13] --- ekr has joined
[13:03:33] --- csp has left
[13:03:55] <ekr> I'm double-booked for DCCP and SIP, but Tom asked me to come in for DTLS/DCCP. Can someone give me a 5 minute warning where it's coming up?
[13:04:40] --- gorryf has joined
[13:05:03] <gorryf> Yes - we can do that.
[13:05:14] <ekr> Thanks. Is it still scheduled towards the end?
[13:05:23] <ekr> (As per the agenda on the web)
[13:05:29] <gorryf> Yes.
[13:05:34] <ekr> Thanks!
[13:05:45] <gorryf> Gorry: Still looking for a jabber scribe for this room.
[13:06:30] <ian.mcdonald> I am in the room
[13:06:48] <ian.mcdonald> But I don't know what a jabber scribe is - do I just have to save the conversation?
[13:07:17] <arjuna_sai> The chair asks if someone in the room is willing to act as "scribe". If no one volunteers, read no further, we're done! The scribe joins the conference room, and types in a running commentary as to what's going on in the room. For example, if a speaker makes a presentation, the scribe types in the URL for the presentation. Simlarly, during question time, a remote participant can type a question into the room and the scribe can pass it on to the speaker
[13:07:23] <gorryf> Yes... Just give the titles of each presenters and tell us if there are questions...
[13:07:39] <gorryf> Is that a volunteer?
[13:07:51] --- kohler has joined
[13:08:06] <kohler> hi all
[13:08:07] <arjuna_sai> I am not sure..I am just pasting it from the IETF site :)
[13:08:09] <ian.mcdonald> I don't feel confident doing that and have no audio to give back into the meeting
[13:08:45] <ian.mcdonald> i mean if somebody asks a question here.
[13:09:06] --- lars has left
[13:10:06] --- falk has joined
[13:10:09] <gorryf> starting the meeting session with Agenda
[13:10:23] <gorryf> Any comments or changes to the Agenda?
[13:10:24] <arjuna_sai> not clear at all..gorry's voice was very clear..
[13:10:37] <gorryf> Better?
[13:10:44] <kohler> a bit
[13:10:48] <arjuna_sai> somewhat better
[13:11:28] --- mallman has joined
[13:11:42] * falk has set the topic to: Overview -- Tom Phelan
[13:11:55] <arjuna_sai> is it possible to increase the sound or is that all?
[13:12:20] --- monden has joined
[13:12:27] <kohler> YES
[13:12:36] <falk> (whiner)
[13:12:40] <arjuna_sai> :)
[13:12:52] <kohler> aaron == brat face
[13:13:09] * falk has set the topic to: Recharter -- Gorry Fairhurst
[13:14:02] --- csp has joined
[13:15:13] --- lars has joined
[13:18:43] <gorryf> Agenda & Milestones done.
[13:18:52] * falk has set the topic to: DCCP Errata -- Tom Phelan
[13:19:06] <gorryf> Active Drafts
[13:20:04] <kohler> a version with editorial changes is sent to errata
[13:20:13] <kohler> (relative to the one sent to list)
[13:20:23] <gorryf> Faster Restart (Sally, 10 min) draft-ietf-dccp-tfrc-faster-restart
[13:20:23] * falk has set the topic to: Faster Restart -- Sally Floyd
[13:20:57] <gorryf> Arjuna, presenting your slides after this - are you in the jabber room?
[13:21:09] <arjuna_sai> yes - I am in here :)
[13:21:11] <falk> can you hear Sally?
[13:21:13] <kohler> yes
[13:21:16] <ian.mcdonald> yes
[13:21:16] <arjuna_sai> yup
[13:22:36] --- jishac has joined
[13:23:38] --- sarolaht has joined
[13:24:05] <falk> Agenda FYI: Datagram Congestion Control Protocol (DCCP) WG WG Chairs: Tom Phelan <tphelan@sonusnet.com> Gorry Fairhurst <gorry@erg.abdn.ac.uk> jabber: dccp@ietf.xmpp.org 1. Agenda Bashing (Gorry, 10 min) * Agenda changes * Election of Scribe for Proceedings * Jabber Scribe 2. Document Status (Tom & Gorry, 10 min) * Updated Charter. * Updated Milestones. * Documents in Last Call - None. * Documents in IESG Review: TCP Friendly Rate Control (TFRC): the Small-Packet (SP) Variant draft-ietf-dccp-tfrc-voip * Documents in RFC Editor Queue - None. * DCCP Errata 3. Active Drafts 3a * Faster Restart (Sally, 10 min) draft-ietf-dccp-tfrc-faster-restart * Issues with faster-restart (A Sathiaseelan via Chair, 5 min) 3b * User Guide and TFRC Media Guide (Tom, 10 min) draft-ietf-dccp-user-guide draft-ietf-dccp-tfrc-media * Feedback from Implementors of DCCP (Ian McDonald via Chair, 10 min) * Feedback from Implementors of RTP/DCCP (Colin, 10 min) 4. Potential New WG Items 4a * DCCP CCID4: TFRC with Small Packets (Sally, 10 min) draft-floyd-ccid4-00 4b * TFRCbis (Sally, 10 min) draft-floyd-rfc3448bis-00 4c * DTLS over DCCP (Tom, 10 min) draft-phelan-dccp-dtls-00 4d * RTP over DCCP (Colin, 10 min) draft-perkins-dccp-rtp 4e * DCCP Mobility (Pasi, 10 min) draft-kohler-dccp-mobility
[13:24:30] --- toyokkie has joined
[13:26:56] <kohler> it's worse, actually -- it reports not 1 packet/RTT, but 1 packet/(the entire length of time including the idle period)
[13:26:58] --- hp has joined
[13:27:19] <arjuna_sai> yes thats right..
[13:27:45] <kohler> not really
[13:27:57] <kohler> OH MY GOD
[13:28:26] <kohler> that's right after the idle period is over, the first feedback packet
[13:28:27] <mallman> dude... you are not looking good these days. you need to take care of yourself
[13:28:34] --- shikob has joined
[13:28:40] <kohler> aaron is cute, dude
[13:28:51] * falk blushes
[13:30:26] * falk has set the topic to: TFRC and Faster Restart -- Gorry for Arjuna
[13:33:40] <kohler> So comfort noise may be a problem for Sally's TFRCbis fix for low-receive-rate-after-idle.
[13:33:53] <kohler> The problem Gorry is talking about now, I don't think is a problem really.
[13:34:05] <kohler> Idle means "idle at the transport".
[13:34:07] --- hp has left: Computer went to sleep
[13:35:02] <kohler> My belief is the changes in the current draft will handle this case well.
[13:35:37] <kohler> what is the milestone again?
[13:36:00] <gorryf> Who volunteers to simulate faster restart?
[13:36:11] <arjuna_sai> I can do it..
[13:36:15] <kohler> Lovely!
[13:36:19] <arjuna_sai> I am currently working on it as well..
[13:36:21] <kohler> Happy to work with you, Arjuna.
[13:36:31] <arjuna_sai> same here kohler thanks :)
[13:36:37] <kohler> Dec 2006
[13:36:44] <arjuna_sai> yeap sure..
[13:37:12] <falk> So, Eddie, Arjuna: do you think you can submit some results by next IETF?
[13:37:15] <gorryf> User Guide (Tom, 10 min) draft-ietf-dccp-user-guide
[13:37:24] <arjuna_sai> yes..
[13:37:30] <kohler> then yes!
[13:37:37] * falk has set the topic to: User Guide -- Tom Phelan
[13:37:44] <arjuna_sai> I have the implementation..tests need to be done.. so yes..
[13:37:56] <kohler> Arjuna, do you agree with me that the current draft handles your "Example Scenario" with buffered data?
[13:38:25] <arjuna_sai> I am still not sure..:(..I need to check this again..
[13:38:48] <arjuna_sai> I shall mail you about this..is that fine?
[13:38:58] <kohler> yes. feel free to mail multiple times
[13:39:02] <arjuna_sai> sure thanks :)
[13:43:29] --- ekr has left: Logged out
[13:44:05] <gorryf> Feedback from Implementors of DCCP (Ian McDonald via Chair, 10 min)
[13:44:29] <ian.mcdonald> I'm watching if any questions as presented
[13:44:32] * falk has set the topic to: Feedback from Implementors of DCCP -- Ian McDonald via Gorry
[13:45:46] --- admcd has joined
[13:47:48] <falk> Ian: what is the phrase at the bottom of your slides?
[13:48:02] <kohler> it's prob Maori
[13:48:44] <ian.mcdonald> it is maori
[13:48:57] <ian.mcdonald> just means university of waikato i think
[13:49:38] <falk> ah. i hoped it was something like "The cool place to be in Waikato"
[13:49:55] <ian.mcdonald> High speed = 100 mbits/sec
[13:50:00] <ian.mcdonald> low latency = 1 or 2 ms
[13:50:12] --- ekr has joined
[13:50:12] <ian.mcdonald> works directly node to node but breaks if router in middle
[13:50:23] <kohler> "breaks"?
[13:50:36] <ian.mcdonald> keeps on having to resync and slows rate dramatically
[13:50:47] <kohler> so the protocol doesn't break
[13:50:54] <ian.mcdonald> protocol doesn't break no
[13:50:57] <kohler> This is not a protocol problem
[13:50:59] <ian.mcdonald> performance does
[13:51:03] <mallman> crap - kohler has an ugly twin, too... who knew?
[13:51:11] <ian.mcdonald> no - but I'm wondering if default is appropriate
[13:51:14] <kohler> Default is right, for security reasons
[13:51:35] <ian.mcdonald> but if this still proves a problem it will be bad perfomance on a LAN/MAN
[13:51:42] <kohler> We might want to argue on the list about how implementations can be given guidance on how to update the sequence window automatically
[13:51:57] <ian.mcdonald> ok
[13:52:06] <kohler> I would expect, for example, that when an implementation notices that the sequence window is getting a bit tight, automatically you should raise it by a factor of two
[13:52:13] * falk has set the topic to: Feedback from Implementors of DCCP -- Mark Handley for Andrea Bittau
[13:52:24] <gorryf> Mark Handley talks on CCID
[13:52:34] <kohler> slides?
[13:52:41] --- uoa has joined
[13:52:41] <falk> on Mark's computer
[13:52:49] <falk> and in Gorry's inbox
[13:52:52] <ian.mcdonald> as a note also on compliance check on RFC what I meant here was verify we have implemented it all - which we haven't - but we haven't made a list of what we haven't done yet...
[13:52:52] <arjuna_sai> :)
[13:54:36] <kohler> Ian: Right. We prob have a while to go first... For instance, we should maybe develop the sequence window guidance before worrying about full RFC compliance.
[13:54:56] <ian.mcdonald> yes i agree
[13:56:23] <kohler> SOCK_DGRAM_CONGESTION_CONTROLLED_BUT_HONESTLY_THIS_IS_MORE_GENERAL_THAN_DCCP
[13:56:45] <csp> :-)
[13:57:04] <ian.mcdonald> SOCK_UNRELIABLE_SESSION?
[13:57:20] <csp> SOCK_UNRELIABLE_STREAM?
[13:57:24] <kohler> Re: service code: This was extensively discussed on linux lists
[13:57:30] <ian.mcdonald> that's what I meant :-)
[13:57:38] <ian.mcdonald> It does have socket buffers now
[13:57:41] <ian.mcdonald> just not merged
[13:58:10] <ian.mcdonald> i'm putting memory limits on the queue etc as now
[13:58:23] <kohler> There is the experimental API, too, that Junwen and I worked on
[13:59:05] <ian.mcdonald> you can set the socket buffer size using socket options
[14:00:02] <csp> Being able to query how much data remains in the send queue would be useful. Is that possible?
[14:00:17] <ian.mcdonald> not yet but I certainly plan to implement that
[14:00:18] <kohler> Some people wanted the service code to be optional
[14:00:21] <kohler> I argued against
[14:00:43] <kohler> That's why we added a service code of 0xFFFFFFFF (invalid) to the service code registry
[14:00:50] <kohler> Currently a socket starts with 0xFFFFFFFF
[14:01:02] <ian.mcdonald> ccid2 and ccid3 do share code
[14:01:10] <ian.mcdonald> they are both in linux kernel
[14:01:17] <kohler> so that the code can detect the lack of a service code setting
[14:01:22] <kohler> The alternative would be to start with 0
[14:01:48] <falk> IF YOU WANT ME TO CHANNEL YOUR COMMENT TO THE MIC, USE THE TAG "FOR ROOM"
[14:02:04] <kohler> ok
[14:04:25] <gorryf> TFRC Media Guide (Tom, 5 min) - done already
[14:04:35] <gorryf> Feedback from Implementors of RTP/DCCP (Colin, 10 min)
[14:05:23] * falk has set the topic to: Feedback from Implementors of RTP/DCCP -- Colin Perkins
[14:09:34] <ian.mcdonald> FOR ROOM: After reading the slides I wonder if some better results for low latency links could be made just by altering timer tick. Old default is 10 msecs and this can vary on newer kernels - 4 msecs or 1 msecs. Can correspond directly if you want.
[14:10:15] --- lars has left
[14:11:51] <kohler> FOR ROOM: Agree with Ian on timer ticks. All of this will apply much less to the final kernel implementation -- assuming there is a kernel queue of some kind!
[14:12:41] <ian.mcdonald> FOR ROOM: I should add you can set the timer ticks to whatever you want with kernel
[14:12:52] <kohler> ian - not clear they are using linux?
[14:13:01] <arjuna_sai> who was that?
[14:13:06] <falk> Mark Handley
[14:13:09] <arjuna_sai> ok..
[14:13:17] <ian.mcdonald> kohler: ok
[14:15:12] <kohler> FOR ROOM: I think many of these problems will disappear with a kernel implementation, with a queue.
[14:15:18] <gorryf> Mark asked would the problems dissapear in kernel land
[14:15:24] <ian.mcdonald> FOR ROOM: I'm running CCID3 down to usecs not msecs with very good accuracy
[14:15:27] <arjuna_sai> isnt that 4 RTT for the nofeedback timer to expire? rather than 2 RTT?
[14:15:35] <gorryf> Matt Mathis talking
[14:15:37] <ian.mcdonald> FOR ROOM: That's in the Linux kernel
[14:15:43] <arjuna_sai> ho ok..
[14:15:57] <kohler> gorry: i understand -- i just think colin is saying "maybe" they'll go away, yes maybe but I think "most probably"
[14:16:19] <kohler> "a little" => "probably a lot"
[14:16:24] <ian.mcdonald> FOR ROOM: We track RTT in usecs as we got exactly these sort of issues when using msecs. DCCP is first protocol to do this in Linux
[14:17:31] <kohler> FOR ROOM: What can we do to help Colin and his students start doing this in kernel?
[14:17:54] <kohler> Arjuna: It should be 4RTT yes
[14:18:49] <ian.mcdonald> Happy to assist other people implementing kernel
[14:19:12] --- mallman has left: Computer went to sleep
[14:19:18] <arjuna_sai> thanks kohler :)..because the slide says 2*RTT...thats y I asked :)..
[14:20:46] <kohler> FOR ROOM: Scheduling problems are often helped by some form of buffering. What is the acceptable buffer here, if any?
[14:22:33] <kohler> those artifacts are pretty
[14:22:40] <kohler> it's art
[14:23:46] <arjuna_sai> how about using quickstart?
[14:23:58] <kohler> arjuna: that's a long way out
[14:24:25] <arjuna_sai> long way out?
[14:24:48] <kohler> quickstart (as opposed to faster restart) requires router changes, at least as i understand it
[14:24:59] <falk> therefore, won't be available for a while
[14:25:16] <arjuna_sai> ok that's a valid point :)
[14:25:40] --- mallman has joined
[14:25:43] <gorryf> Conclusions from CP - if you have last questions say...
[14:26:05] <kohler> FOR ROOM: My instinct is that even some of the later problems might become easier with a kernel implementation.
[14:26:19] --- mallman has left: Computer went to sleep
[14:27:00] <kohler> FOR ROOM: I do not think we should write TFRC media guide based on user level implementations, where it is simply not clear which issues are coming from the user level implementations and which are not.
[14:28:04] <kohler> FOR ROOM: Kernel impl just makes buffering easier. Not all problems will go away clearly... don't know which will and which won't.
[14:28:36] <kohler> So guidance says, Don't do it at userlevel?
[14:29:15] <kohler> YES
[14:29:22] <kohler> (do it with a kernel impl)
[14:30:33] * falk steps away
[14:32:43] <gorryf> * Next: DCCP CCID4: TFRC with Small Packets (Sally, 10 min) draft-floyd-ccid4-00
[14:38:06] <arjuna_sai> Is there a limit on how big these small packets are?
[14:43:22] <gorryf> Moving quickly on: TFRCbis (Sally, 10 min) draft-floyd-rfc3448bis-00
[14:44:11] * falk has set the topic to: DTLS over DCCP -- Tom
[14:45:53] --- admcd has left
[14:47:45] --- mallman has joined
[14:51:09] --- monden has left
[14:52:00] <gorryf> Moving quickly on: TFRCbis (Sally, 10 min) draft-floyd-rfc3448bis-00
[14:53:04] --- hp has joined
[14:53:12] * falk has set the topic to: TFRCbis (Sally, 10 min) draft-floyd-rfc3448bis-00
[14:56:33] <falk> any jabber humming?
[14:56:37] <kohler> HUMM on pro
[14:56:48] <falk> fyi: all hums were pro
[14:57:30] <gorryf> * RTP over DCCP (Colin)
[14:59:22] --- hp has left: Computer went to sleep
[15:03:09] <gorryf> Next:Going to ask for people who have read and ask for WG adoption.
[15:03:22] --- mallman has left: Computer went to sleep
[15:03:45] --- jishac has left
[15:04:20] --- jishac has joined
[15:05:35] --- falk has left
[15:05:39] <sarolaht> Eddie- we are skipping mobility presentation. ok?
[15:05:50] --- falk@jabber.isi.edu has joined
[15:05:56] <sarolaht> new people floating to room...
[15:06:17] --- csp has left
[15:06:38] <kohler> hum on more time, but also ok with progressing
[15:07:03] <kohler> FOR ROOM: THAT IS A LIE
[15:07:06] <kohler> draft was updated
[15:07:19] <falk@jabber.isi.edu> well out of time either way
[15:07:23] <kohler> yeah
[15:07:23] <falk@jabber.isi.edu> next group has arrived
[15:07:26] <kohler> just sayin'
[15:07:28] <kohler> bye all!
[15:07:37] --- shikob has left
[15:07:45] <ian.mcdonald> bye
[15:07:46] --- sarolaht has left
[15:07:47] <falk@jabber.isi.edu> ttfn
[15:07:47] --- kohler has left
[15:07:52] --- toyokkie has left
[15:07:52] --- ian.mcdonald has left
[15:07:53] --- falk@jabber.isi.edu has left: Logged out
[15:08:12] --- gorryf has left
[15:08:52] --- uoa has left
[15:10:00] --- arjuna_sai has left: Replaced by new connection
[15:11:50] --- ekr has left
[15:18:32] --- jishac has left
[15:37:30] --- mallman has joined
[15:37:42] --- mallman has left: Computer went to sleep
[15:45:09] --- mallman has joined
[15:45:22] --- mallman has left
[17:15:07] --- LOGGING STARTED