[09:51:20] --- sventers has joined [09:52:23] --- Danny has joined [09:52:29] --- kodonog has joined [09:52:42] hi Stewart [09:52:56] and Karen [09:53:05] Howdy all [09:53:20] Hi Danny and Stuart [09:57:42] --- lj324o230ar has joined [10:00:58] --- Stewart Bryant has joined [10:02:26] --- adam.wertheimer.telcordia has joined [10:02:48] all presentations are on the meeting material site [10:03:49] --- Tony Li has joined [10:04:04] Hi all [10:04:13] Luca to do jabber [10:04:14] --- lllmartini has joined [10:04:47] We're now chartered [10:05:00] See the IETF page. Ignore unofficial page [10:05:21] 3.5 drafts [10:05:24] in queue [10:05:26] --- Mark Townsley has joined [10:06:05] chairs wrote most of the drafts , and are looking for folks to take over [10:06:19] ( to avoid admin awkardness ) [10:09:14] --- jmhalpern has joined [10:09:32] microphone : I think we should start to thin kwhat the technology will do before doing the MIB [10:09:59] That was kurtis [10:10:20] work items: we do ntp extension here in tictoc [10:10:55] and we do 1588 profile here == specific parameters/extension are done in tictoc [10:11:27] ITu also working on 1588 profile , hence we need coordination. [10:12:32] mark is baching the agenda [10:12:41] --- jmhalpern has left [10:12:43] let's start talking about requirements [10:14:12] definition of "on-path" support [10:18:19] Requirement: GSM needs .05ppm accuracy [10:18:36] SDH provide 50ppm [10:18:37] Base station ... [10:19:27] Phase sync requirement for 3G +- 3us [10:19:48] Luca comment: I can easily get 250ns ..... on large MPLS network [10:20:01] Currently using GPS as reference clock [10:20:22] Don't like it, as it's inflexible, would prefer distributed clocking [10:23:42] Pointless filler material... [10:24:01] --- john.zhao has joined [10:24:07] :-) [10:25:30] More filler... [10:26:09] All of this boils down to: we need to run over IP. [10:26:15] In the ITU this is considered revolutionary - but we have to bring them with us [10:26:20] or mpls [10:26:25] yes - but not everybody is familiar with RAN [10:26:39] Have you considered running applications directly over MPLS? No thank you. [10:26:45] --- markgre has joined [10:26:57] Luca, you get 250ns, with what sort of packet loads distrubing things [10:26:58] Therefore it runs over IP, which in turn may be over IP. [10:27:02] no , IP is great for applications [10:27:18] And you get MPLS for free... [10:27:26] no matter what packet load .... [10:27:42] not ethat worse condition is at 96% to 99% load on a link [10:28:16] I would like to see us move to reqs for the protocol we are going to build here soon. We've beat the use cases to death in the BoFs already... [10:28:20] and note that that was not for time sync transmission , only frequency [10:28:24] If you played diffserv games or used your routers as part of your NTP mesh, the loading wouldn't matter. [10:28:43] diffserv ... and the load matters ... [10:28:50] --- kurtis@jabber.psg.com has joined [10:28:59] what kind of accuracy is he looking for? [10:28:59] even a real time queue is not real time .... [10:29:01] Not if NTP gets priority queuing... [10:29:15] True, I don't mean that there's zero jitter. [10:30:02] HW designers put unprioritized FIFOs in the path .... [10:30:32] SDH requirement in the order of 10^-11 [10:30:45] Kurtis is up next [10:31:41] Kurtis gets to use his first name on the draft name.. Does that mean he is like Cher or Madonna? [10:31:57] Funny, he doesn't look as good as either of them. [10:32:08] sonny [10:32:11] depends [10:32:17] --- juampe has joined [10:32:37] (apologies for forgetting this wasn't bad-attitude) [10:32:38] 1PS ??? WOW! [10:33:20] Yeah, I'm surprised too. Coming from Peter I was expecting 10^-17 [10:33:42] Mark: backwards compatibility in charter [10:34:21] NTP will reply using the version requested [10:34:22] we agree that we will have backward compatibiity [10:34:59] Dave Oran: nods, [10:35:27] Dave: something is potentially missing : notion of higher precision time protocol [10:35:34] running as sub-strate [10:35:37] in an ntp system [10:36:19] charter should allow a sub sytem that then can export time to ntp ( or using NTP ) [10:37:22] Peter: we are trying to talk a protocol to transports tiem across the internet ... all we are saing is that we want to use the same port , and we must reply to requests that use the old version of ntp [10:37:27] What slide set is Kurtis talking from? The last presenter (China Mobile) was using tictoc-6. [10:37:47] Where does 1ps requirement come from? [10:37:48] slide title is " backward compatibility" [10:38:33] Requirements for a next generation Internet Time Protocol - tictoc-7 - slide 3 [10:38:39] he is talking from tictoc-7.pdf [10:38:50] yes [10:39:29] what about mode 6 and 7? [10:39:42] which are management modes [10:40:26] the real point was backward compatibility is a must - they started delving into the details of what that meant and mark moved them back to the general rqt [10:42:11] Mostly this seems like a rathole [10:42:30] 1ps = 1/100 bit time at 10G, seems pretty specialized for quite a while in the future [10:43:03] where did this requrement come from ? [10:43:33] who knows. Why would anyone want to do this? It causes problem [10:43:44] I don't think it's a requirement, I think it's what they're proposing. [10:44:07] Why would anyone want to keep maser's in their basement? [10:44:52] disagree [10:45:08] Disagree with what? [10:45:10] with what? [10:45:26] Seems like if client needs to calculate dispersion, then adding server is not much more. [10:45:27] separation between end client and client/server [10:45:27] Peter does ! [10:46:46] --- steve has joined [10:47:02] Stewart /Yaakov - this discussion is more relevant than my presentation and you can take time from it if necessary [10:47:32] tks [10:49:00] Peter: we need to support clock sync on Mars ... [10:49:47] API needs to include : freq, phase offset , clock source characteristics .. .etc.. [10:50:04] The same problems apply elsewhere - to aircraft I think [10:51:21] Transfer of time to a different inertial reference is a requirement? [10:52:16] why not ? [10:53:18] I don't think the definitions for TAI or UTC work in that case. [10:53:30] correct [10:53:37] tictoc meeting cannot run over , by definition ;-) [10:54:22] just overclock ;) [10:54:37] no enought heat dissipation in this room [10:54:44] Who's up next? [10:54:54] Greg Dowd [10:54:55] Sorry, Greg Dowd [10:55:09] they are going out of order [10:55:16] why? [10:55:33] Uh... mark asked them to focus on requirements [10:55:38] Sorry we were asked to do requirements first [10:55:41] They are using Greg's preso to transition to 1588 [10:55:56] gap analysis? [10:56:01] This is not quite requirements but there some good design hints [10:56:04] yet gap analysis [10:58:25] --- chip has joined [11:00:02] --- john.zhao has left: Replaced by new connection. [11:00:02] --- john.zhao has joined [11:02:17] --- john.zhao has left: Replaced by new connection. [11:02:18] --- john.zhao has joined [11:03:44] --- john.zhao has left: Replaced by new connection. [11:03:45] --- john.zhao has joined [11:04:20] --- Mark Townsley has left [11:04:43] --- Mark Townsley has joined [11:05:36] BMC is best master clock [11:10:08] --- chip has left [11:10:15] --- john.zhao has left: Replaced by new connection. [11:10:16] --- john.zhao has joined [11:10:19] --- Chip Sharp has joined [11:11:17] --- Ed J. has joined [11:11:38] --- john.zhao has left: Replaced by new connection. [11:11:39] --- john.zhao has joined [11:13:40] --- john.zhao has left: Lost connection [11:13:51] --- Tony Li has left: Computer went to sleep [11:14:00] hoho tictoc is not keeping time ... [11:14:52] --- john.zhao has joined [11:16:37] Decreased polling interval? [11:16:46] no [11:17:08] So we need the ability to poll slower? [11:18:54] polling too frequently does not provide any real benefit and you overwelm the servers unnecessarily [11:19:04] It's a common mistake that people make [11:19:18] it's called oversampling [11:19:45] yes - the important part is to know what to consider a valid sample [11:19:57] Doesn't oversampling increase my odds of getting low dispersion measurements? [11:20:40] Semtech have some results that show that 30samples per sec is optimum over existing nets for freq xfer [11:20:46] and the odds of becoming a victom of network wander [11:21:24] no they groom over a long period of time - they need more lucky packets [11:22:14] Which presentation/page? [11:22:32] Laurent montini gen items 2/2 [11:22:39] tictoc-2 (Laurent) [11:22:44] yes [11:24:39] tranparent clocks 3 [11:24:45] sorry 2 [11:27:47] --- john.zhao has left: Disconnected. [11:27:51] --- john.zhao has joined [11:29:19] 1588 telecom profile [11:30:21] tictoc-4 [11:30:57] I think this presentation is bit premature... [11:33:15] directly over ethernet ? [11:33:31] ethernet is fery short range .... how is this usefull ? [11:35:28] Maybe they are thinking about Carrier Ethernet (range of 10s or 100s of Kms) [11:35:39] likely [11:36:09] the time of large bridge networks is over .... I meant to say that ethernet is a one network hop only proposition [11:36:21] it is works over IP it could be a lot more useful [11:37:14] --- harlan has joined [11:38:24] legacy = not IEEE1588 [11:38:41] or the Internet [11:42:07] --- john.zhao has left: Lost connection [11:42:27] Long line forms at mic... [11:42:30] HAving a real application is good for a reality check for a generic protocol design. [11:42:57] Kurtis says IETF works better when we start with a problem, and then develops a solution [11:43:20] We just have to be careful to not get too focused on solving just one application. [11:44:35] It's a fundamental design change to make the accuracy better than a bit time. [11:48:15] --- Tony Li has joined [11:48:17] which talk? [11:48:39] tictoc-3? [11:48:53] Silvana - yes - tictoc-3 [11:48:58] Profiles? [11:49:04] she's on slide 3 [11:49:21] is this on Profiles? [11:49:26] yes [11:50:17] Stupid question: why do the mobile operators need such accuracy? [11:50:54] simultaneus sycronized transmission from multiple cells [11:51:16] For handset triangulation? [11:51:18] you receive multiple simultaneus version , and roam along without incident [11:51:19] To triangulate to locate a customer who doesn't have GPS? (Also maybe for handloff?) [11:51:34] nothing to do with triangulation [11:52:00] CDMA depends on the fact that the sequences from different base stations start as the same absolute time to ensure that the different sequences do not interfere (orthogonal). [11:52:21] Ah, thank you. Interference separation. Got it, makes sense. [11:52:31] Generally mobile people want high frequency standards as that drops less calls [11:55:22] it would kill autokey! [11:57:54] Maybe TC is a different trust model, guy in middle is now friend? [11:59:58] Which slide pack? [12:00:11] tictoc-5 [12:00:18] --- john.zhao has joined [12:00:32] --- lllmartini has left [12:01:02] --- kurtis@jabber.psg.com has left [12:01:46] --- Ed J. has left [12:01:53] Mark: Enough use cases, please give us requirements and architecture [12:02:07] --- Mark Townsley has left [12:02:12] Kurtis: Interim? [12:03:28] All done! [12:03:30] --- Tony Li has left [12:04:27] --- lj324o230ar has left [12:04:31] --- adam.wertheimer.telcordia has left [12:05:08] --- harlan has left [12:05:52] --- john.zhao has left: Computer went to sleep [12:06:50] --- sventers has left [12:16:21] --- john.zhao has joined [12:18:19] --- Stewart Bryant has left [12:29:25] --- sventers has joined [12:32:18] --- Chip Sharp has left [12:32:29] --- steve has left [12:33:06] --- john.zhao has left [12:44:38] --- markgre has left [13:08:41] --- lllmartini has joined [13:27:10] --- lllmartini has left [13:39:49] --- kodonog has left: Replaced by new connection [14:07:46] --- Mark Townsley has joined [14:09:14] --- Mark Townsley has left [14:17:55] --- Mark Townsley has joined [14:18:36] --- Mark Townsley has left [14:25:03] --- Stewart Bryant has joined [14:26:25] --- Stewart Bryant has left [14:29:58] --- Stewart Bryant has joined [14:31:49] --- sventers has left [14:39:00] --- Danny has left: Replaced by new connection [14:39:00] --- Danny has joined [14:51:52] --- Danny has left [15:05:28] --- Stewart Bryant has left [15:43:46] --- Stewart Bryant has joined [15:56:09] --- Stewart Bryant has left [16:12:02] --- Stewart Bryant has joined [16:15:23] --- Stewart Bryant has left [18:42:23] --- Danny has joined [18:55:59] --- Danny has left