- Tuesday, July 30, 2013

aqm@jabber.ietf.org

Tuesday, July 30, 2013< ^ >

Room Configuration

Room Occupants

GMT+0
[10:11:28] rscheff joins the room
[10:50:37] rscheff leaves the room: Computer went to sleep
[11:01:36] rscheff joins the room
[11:57:33] rscheff leaves the room: Computer went to sleep
[11:58:18] rscheff joins the room
[12:33:37] rscheff leaves the room: Computer went to sleep
[13:33:05] andrey.uzunov joins the room
[13:33:09] andrey.uzunov leaves the room
[14:40:25] davetaht joins the room
[14:41:44] davetaht leaves the room
[14:42:03] davetaht joins the room
[14:47:40] jgunn joins the room
[14:58:07] rscheff joins the room
[15:00:07] <davetaht> yea, the usual suspects.
[15:00:49] Xavier Marjou joins the room
[15:02:43] <jgunn> Audio seems good
[15:02:45] jlcJohn joins the room
[15:03:58] <jlcJohn> Is anyone else getting "This webpage is not available" for the materials link? (https://datatracker.ietf.org/meeting/87/materials.html#wg-aqm)
[15:04:48] <jgunn> I got the materials OK
[15:05:44] <jgunn> Which slide deck are we on?
[15:06:19] <jgunn> Do we have a jabber sceibe?
[15:06:55] cabo joins the room
[15:11:09] danyork joins the room
[15:11:33] Scott Brim joins the room
[15:13:22] Lars joins the room
[15:14:13] <rscheff> slide-87-aqm-5
[15:14:13] <rscheff> slide 8
[15:14:41] <jgunn> thanks
[15:14:52] <Scott Brim> Janet, do you have audio?
[15:15:05] <jgunn> Yes, I have audio.
[15:15:30] <davetaht> I have great audio... tunneled ipv6, mesh network, to california. Applause.
[15:15:41] Michael Richardson joins the room
[15:18:15] rscheff leaves the room: Replaced by new connection
[15:18:16] rscheff joins the room
[15:19:23] Xavier Marjou leaves the room
[15:20:23] Y(J)S joins the room
[15:21:20] Ted joins the room
[15:21:58] <Ted> Hi Janet
[15:22:17] <Ted> Did you get an answer?  If not, we're on the "Conclusions/Recommendations" slide from Fred's talk.
[15:25:20] <jgunn> Yes, "rscheff" kindly posted the slide numbers.  It was when Fred was talking about "the picture" that I could not tell WHICH picture  he was referring to.
[15:25:23] <jlcJohn> Howcum nobody is objecting to Rec# 2 (ECN)?
[15:25:32] Guillaume Leclanche joins the room
[15:25:58] <jgunn> Hi Ted
[15:27:24] <Ted> Where "hitting you hardest" is apparently "will have largest dwell time in the queue"
[15:27:32] <Scott Brim> yah
[15:27:54] <Ted> Now moving to PIE slide deck; on the title slide now
[15:27:58] <davetaht> "The packet you want to drop is the one that is going to wait the longest". -- yea, what ted just resaid... shoot at the elephants.
[15:28:31] <Ted> Rong Pan speaking representing the author team.
[15:28:42] <Ted> Moving to slide 2 "The Problem of Buffer Bloat"
[15:29:55] <Ted> Now on slide 3 "Control Average Delay"
[15:29:59] <Ted> Reverted to slide 2
[15:30:28] gorryf joins the room
[15:30:52] tuexen joins the room
[15:30:59] Wolfgang Beck joins the room
[15:31:16] <Ted> back to slide 3, "Control Average Delay and Allow Big Burst"
[15:32:24] <Ted> This seems to be an argument for queue segratation
[15:32:36] <Ted> "Water Level in a leaky Bucket" slide, (slide 4)
[15:32:46] Jason Weil joins the room
[15:33:42] <Ted> "Pie slide"  5
[15:33:54] <Ted> Goal #1 slide, slide 6
[15:34:59] <Michael Richardson> is it fair to say that PIE is an automatic RED tuner based upon output queue delay?
[15:35:06] <Ted> Now at Goal #3 Achieving High link utilization
[15:35:56] <davetaht> um, noo... there are a couple new ideas in pie, notably byte sensistive dropping probability and the three tier threshold based on delay, not queue length...
[15:36:44] andrey.uzunov joins the room
[15:36:50] <Ted> Now at "design of PIE", slide 9
[15:37:38] shep joins the room
[15:37:51] <Ted> Now "the block diagram of PIE", slide 10
[15:38:13] andrey.uzunov leaves the room
[15:38:21] <Ted> Now at slide 11, PIE's work update
[15:38:29] timcopley_gmail.com joins the room
[15:39:22] <Ted> Now at "summary and future work", slide 12
[15:40:54] <Ted> Hannes at the mic.
[15:41:30] <shep> Her slides on the screen say "Cisco Confidential" in the lower right hand corner.
[15:41:53] <Ted> Jim Gettys at the mic
[15:42:34] <Ted> Ken Gray at the mic
[15:43:39] <Ted> Fred Baker at the mic
[15:43:42] <davetaht> My version of pie does ecn. linux codel and fq_codel have ecn. The theoretical problems with ecn on the wild and open internet remain.
[15:44:07] <Ted> Andrew McGregor going forward for "QoS and AQM"
[15:44:18] <Ted> (Did no catch speaker's name)
[15:44:33] <Ted> Fred Baker goes back to mic
[15:45:15] <Ted> Slide now fq_codel Status, with Jim Getty's name but Andrew up.
[15:45:33] <gorryf> Who was at the Mic?
[15:45:36] <Ted> Rong Pan back the mic
[15:45:47] <Ted> er, back at the mic.
[15:46:26] <Ted> Mitch(?) from IBM research speaking
[15:47:14] <Ted> Andrew answers.
[15:47:23] <Ted> Now on slide 2
[15:48:30] <Ted> I thought I understood airport security queues, but I don't see the relationship here…..
[15:49:24] <Ted> Now on slide 3 "fq_codel works"
[15:50:27] <shep> (He's on the FQ_CoDel Status  slides, not his slides. )
[15:50:33] <Ted> Slide 4 "implentation status"
[15:50:55] <Ted> Back a slide in response to Fred's request
[15:52:14] davetaht leaves the room
[15:52:27] <Ted> Jim Gettys responding
[15:52:36] <Ted> Hannes speaking
[15:53:20] dave.taht joins the room
[15:53:23] <dave.taht> or mac addr
[15:53:44] <Ted> Moving back to implementaiton status
[15:53:49] Guillaume Leclanche leaves the room
[15:53:52] <Ted> "Linux Implementation"
[15:54:21] dave.taht leaves the room
[15:54:27] <Ted> Now on "Next Steps" slide
[15:55:07] Jason Weil leaves the room
[15:56:28] <Ted> Now on "Qos and AQM" slide deck
[15:56:30] <shep> (Now Andrew has switched to his slides.)
[15:56:43] andrey.uzunov joins the room
[15:56:47] <Ted> "It's about constraints" slide
[15:57:31] dave.taht joins the room
[15:57:46] <Ted> "What behaviours make sense" slide
[15:58:55] <Ted> fq_codel slide now up
[15:59:05] <Ted> "Some implicit flow behavior" sub heading
[16:00:15] <Ted> Limitations on deduction
[16:01:20] <Ted> Layers below IP slide
[16:03:12] tuexen leaves the room
[16:03:25] <Ted> Now on "shallow bufering"
[16:05:20] <Ted> Criteria for success
[16:06:36] <shep> Ted Hardie at the mic
[16:07:49] <shep> Jim Gettys at the mic
[16:08:06] <shep> Ted Hardie at the mic
[16:08:58] <Ted> Stuart Cheshire at the mic
[16:08:58] <shep> Stewart Cheshire at the mic
[16:09:17] <shep> oops, I yield to Ted
[16:09:46] <Ted> No, no we can be independent queues and multipath the data!
[16:09:48] <dave.taht> the network video (netflix) problem is somewhat exaberated by the mid-short RTTs involved in co-locating that stuff in the ISP data center, so you end up with RTTs in the 2-20ms range which is a grey area between the data center and the world wide internet.
[16:10:08] <dave.taht> +10 on stuart cheshire's comments.
[16:10:55] <Ted> Separation gives us responsiveness under load
[16:11:08] <dave.taht> +11 on his last comment.
[16:11:24] <Ted> Jim back up
[16:11:39] <dave.taht> funding has been something of a problem. :understatement:
[16:12:40] <Scott Brim> who else is skipping the social event tonight?
[16:12:44] <Ted> Andrew's final thing:  one algorithm or many?  fq_codel makes a big difference in some places, but there are other places where it can't be deployed.
[16:13:02] <Ted> So we have to many, because we can't have the same thing everywhere.
[16:13:30] <Ted> Chairs skipping the "compare and contrast" section, in order to get to the Bof questions
[16:14:20] <Ted> Now on "BoF Questions" slide
[16:14:26] <dave.taht> how many people are in the physical room?
[16:14:53] <Ted> It's a big room; maybe a 100?
[16:14:55] <dave.taht> hmmmmmmm
[16:15:03] <dave.taht> wow
[16:15:38] <shep> way more than one hundred by my quick estimate.
[16:15:40] <Ted> That was David ; Andrew responds
[16:16:14] <Ted> David promises to have words with him.
[16:16:27] <shep> 12 rows with 15 or more people per row, is my estimate, so 180 or more people.
[16:16:41] <dave.taht> double wow.
[16:16:46] <Ted> Thanks for estimating; we'll eventually get blue sheets.
[16:17:07] <shep> some rows have more than 20 people, so could easily be over 200 people in the room.
[16:17:24] dave.taht helped get to this point on nearly zero budget while sleeping on couches and in a yurt.
[16:17:36] <dave.taht> Can I stop now?
[16:17:38] <Ted> A yurt?
[16:18:02] <dave.taht> http://www.bufferbloat.net/projects/cerowrt/wiki/Yurtlab
[16:18:08] <Scott Brim> You had a yurt?  In my day we had to sleep outside on horse blankets
[16:18:22] <dave.taht> heheheh.
[16:18:40] <dave.taht> The name of darn group remains an issue.
[16:18:54] <Ted> You can ask to name it "Yurt"
[16:18:58] <Scott Brim> QUEUE: Queue Utilisation for Enhanced User Experience
[16:19:00] <Scott Brim> from Grenville
[16:19:06] <Scott Brim> easy winner
[16:19:22] <Y(J)S> Scott - I agree !
[16:19:34] <dave.taht> Queue. Not bad.
[16:19:54] <dave.taht> someone take the mic for nailing the name.
[16:21:48] <shep> did blue sheets circulate in this BOF?   I don't remember.
[16:22:17] <dave.taht> goforthehum
[16:22:29] <Scott Brim> .com
[16:23:00] <Y(J)S> Is there a jabber to hum conversion app ?
[16:24:04] <jgunn> How can you "recharter" before the WG forms?
[16:24:21] dave.taht leaves the room
[16:24:52] dave.taht joins the room
[16:24:57] <Scott Brim> the third question shouldn't exist.  If we decide to exclude flow isolation, then in the future we might decide to recharter
[16:26:14] dave.taht waits for the vote
[16:26:40] <Scott Brim> we are descending into confusion.  What question actually matters.
[16:27:03] <jlcJohn> ++Scott
[16:28:23] <gorryf> Time to cut the discussion and go to the social -- or decide to form a WG?
[16:28:42] <dave.taht> need a wg name, and a hum to include packet scheduling or not...
[16:28:51] <Y(J)S> Thre is still a whole minute ...
[16:28:59] <gorryf> :-)
[16:29:04] <dave.taht> 100 people in the room to me = need working group. :)
[16:29:31] <Michael Richardson> 100 people with tomatoes can make ketchup!
[16:29:40] <Y(J)S> BTW, maybe Queue Unleashing for Enhanced User Experience
[16:29:48] <Ted> If they have vinegar, salt sugar, and selected spices!
[16:30:41] <Scott Brim> OK :)
[16:33:49] <Ted> People are not going to the social, it appears.
[16:35:16] Lars leaves the room
[16:35:16] Lars joins the room
[16:35:16] Lars leaves the room
[16:35:18] <Michael Richardson> AQM WG should be formed.
[16:35:20] Scott Brim leaves the room
[16:35:24] shep leaves the room
[16:35:27] <dave.taht> groovy
[16:35:32] <timcopley_gmail.com> humm
[16:35:39] Wolfgang Beck leaves the room
[16:35:51] <Y(J)S> We alerady discussed charter items, so I guess there must be a WG ...
[16:35:56] cabo leaves the room
[16:36:02] <Y(J)S> Is it done ?
[16:36:14] <dave.taht> Yea. I think we have a wg.
[16:36:15] andrey.uzunov leaves the room
[16:36:18] <timcopley_gmail.com> I think Audio is about 30 seconds behind the group.
[16:36:44] <Y(J)S> The TCP audio has about 10 - 15 seconds delay
[16:37:05] Ted leaves the room
[16:37:07] <Michael Richardson> yeah, you can't hum in jabber, you need to be explicit.
[16:37:08] <dave.taht> I fear I was humming out of sync then. But everybody knows my opinion.
[16:37:27] <dave.taht> how loud was the hum for packet scheduling?
[16:37:36] <timcopley_gmail.com> I probably hummed out of sync as well
[16:37:52] Y(J)S leaves the room
[16:38:00] gorryf leaves the room
[16:38:04] <dave.taht> Unlike other members at the mic I welcome new attempts at better packet scheduling. There's this guy doing weird stuff with gpus...
[16:38:19] <jgunn> goodbye
[16:38:22] jgunn leaves the room
[16:38:24] <dave.taht> adios!
[16:38:52] Michael Richardson leaves the room
[16:39:21] rscheff leaves the room: Computer went to sleep
[16:44:44] Lars joins the room
[16:45:43] <dave.taht> hi lars, did you ever get anywhere your bsd implementation?
[16:46:40] <Lars> i didn't find the right student to start it yet
[16:47:33] Lars leaves the room
[17:14:06] Wolfgang Beck joins the room
[17:16:16] danyork leaves the room
[17:37:32] timcopley_gmail.com leaves the room
[17:49:44] jessian joins the room
[17:49:50] jessian leaves the room
[18:05:02] Wolfgang Beck leaves the room
[18:09:19] Wolfgang Beck joins the room
[18:09:52] Wolfgang Beck leaves the room
[18:10:24] dave.taht leaves the room
[18:10:25] Wolfgang Beck joins the room
[18:11:00] dave.taht joins the room
[18:15:04] dave.taht leaves the room
[18:16:20] dave.taht joins the room
[18:17:33] dave.taht leaves the room
[18:18:29] dave.taht joins the room
[18:31:55] Wolfgang Beck leaves the room
[19:01:01] dave.taht leaves the room
[19:26:32] andrey.uzunov joins the room
[19:27:08] andrey.uzunov leaves the room
[19:58:12] Wolfgang Beck joins the room
[20:17:05] Wolfgang Beck leaves the room
[20:17:50] Wolfgang Beck joins the room
[20:45:40] Wolfgang Beck leaves the room
[20:45:56] Wolfgang Beck joins the room
[20:53:05] Wolfgang Beck leaves the room
[20:53:41] Wolfgang Beck joins the room
[21:35:16] Wolfgang Beck leaves the room