[13:41:59] --- rik wade has joined
[14:52:40] --- dinakar has joined
[14:52:48] --- dinakar has left
[14:54:47] --- dinakar has joined
[14:54:55] --- dinakar has left
[14:55:41] --- dinakar has joined
[14:55:43] --- mallman has joined
[14:59:15] <rik wade> Is there a scribe for this session? (I'm not on-site)
[15:04:15] --- dinakar has left
[15:09:12] --- dinakar has joined
[15:09:15] --- dinakar has left
[15:15:04] --- mallman has left
[15:22:23] --- rik wade has left: Replaced by new connection
[15:22:24] --- rik wade has joined
[15:23:08] --- taiji has joined
[15:23:30] --- taiji has left
[15:28:35] --- mankin has joined
[15:28:46] <mankin> looks like no scribe
[15:28:51] <rik wade> no
[15:29:13] <mankin> I'm their AD, but not in the room now - I'll ask them to get a scribe when I go over
[15:29:28] <rik wade> great, thanks.
[15:29:37] <mankin> may be low-key, but...
[15:36:42] --- sleinen has joined
[15:37:39] --- mankin has left: Disconnected
[15:37:43] --- mattz has joined
[15:37:50] <sleinen> I only just managed to join jabber from the meeting room - will try to transcribe
[15:38:01] <mattz> yay! thanks simon. I'll also try to help
[15:38:01] <sleinen> Anura Jayasumana http://www.ietf.org/internet-drafts/draft-jayasumana-reorder-density-03.txt
[15:38:36] <mattz> btw, this session is supposed to be both multi- and uni-cast.
[15:38:42] <sleinen> shows different packet sequences
[15:38:50] <sleinen> 1,2, 8, 3,4,5,6,7,9,10,11
[15:38:58] <sleinen> packet 8 is probably the one that got reordered
[15:39:34] <sleinen> not always easy to pick a specific packet that is early or late
[15:39:44] <sleinen> Reordering Density tries to work around this
[15:40:01] <sleinen> Definition: reorder event r(m, dm)
[15:40:34] <sleinen> Use displacement measure (d_m)
[15:40:54] <sleinen> Draw histogram earliness/lateness (d_m) vs. reordering density (on y axis)
[15:41:06] <sleinen> Ex. 2 with packet duplication
[15:41:38] <sleinen> threshold (for displacement?) keeps mem. requirements bounded
[15:42:02] <sleinen> Ex. 3: with packet loss
[15:42:10] <sleinen> 1,3,4,2,6,7,8 (packet 5 is lost)
[15:42:17] --- jishac has joined
[15:42:24] <sleinen> Assignment of receive_index values
[15:42:31] <sleinen> ...
[15:42:44] <sleinen> Result: RD orthogonal to duplication and loss.
[15:42:57] <sleinen> Two detection methods for loss
[15:43:01] <sleinen> * Go-back DT
[15:43:09] <sleinen> * Stay-back DT
[15:43:42] <sleinen> Salient features of RD
[15:43:58] <sleinen> * RD approaches PDF of packet displacement as length of sequence increases
[15:44:06] <sleinen> * orthogonal to loss & dup
[15:44:22] <sleinen> * robust in presence of losses, burst lsos, sense of proportionality
[15:44:31] <sleinen> * useful: can be used e.g. for setting TCP dupthresh
[15:44:45] <sleinen> * Complexity: space O(DT), computation O(N)
[15:44:54] <sleinen>
[15:45:03] <sleinen> Reorder Response of a network [J]:
[15:45:11] <sleinen> RD at the output of the subnet for a sequence of in-order packets
[15:45:31] <sleinen> * a reorder response exists for a network in which packet reordering is not dependent on packet seq no
[15:45:47] <sleinen> * reorder response depends on distribution of inter-packet gaps and network operating condition
[15:45:57] <sleinen> * Ref. http://www.cnrl.colostate.edu
[15:46:19] <sleinen>
[15:46:24] <sleinen> RD and Reorder Response
[15:46:36] <sleinen> Example of two concatenated networks
[15:46:56] <sleinen> A -> Net-X -> B -> Net-Y -> C
[15:47:04] <mattz> [ http://videolab.uoregon.edu/events/ietf/ietf60.html or http://esm.cs.cmu.edu for broadcasts ]
[15:47:48] <sleinen> reorder response of composite network J[k] is convolution of J1[k] and J2[k] (J1[k] * J2[k])
[15:48:31] <sleinen> [Yes, please use Audiocast/videocast, it's hard to keep up typing :-]
[15:49:56] <rik wade> sorry, I'm only on ISDN here at the moment ;-)
[15:50:38] <sleinen> Shouldn't that work for the audio-only feed (MP3?)
[15:51:14] <mattz> mallman: kind of understand scheme... but what insight do these numbers give me for network?
[15:51:40] <sleinen> [audio URL for this meeting: http://videolab.uoregon.edu/events/ietf/ietf-ch4.m3u (64kbps MP3)
[15:51:46] <rik wade> sleinen: got it, thanks.
[15:52:12] <mattz> a: can tell when reordering is due to two paths... is working on linking to causes now
[15:52:41] <mattz> ...can understand nature of reordering
[15:52:49] <mattz> mallman: but still trying to understand _meaning_
[15:53:10] <mattz> looking at example on p.24 (for when we can look there)
[15:53:18] <mattz> see reorder w/in 3 packets
[15:53:32] <mattz> look at distribution... from that can gain insight into network issues
[15:54:38] <mattz> al: so real value is measure and cascade results... need more investigation. desirable quality in current IPPM reordering draft... but intuitively can't believe that these distn's are independant of packet spacing
[15:54:57] <mattz> al == al morton
[15:57:01] <mattz> al: would like to harmonize basic definition of reordering (so same as what current draft is, and what others use)
[15:57:36] <mattz> ans: look at system theory. not predict what happens with any given packets, but statistical nature
[15:57:50] <mattz> think that addresses first isue (independance of spacing)
[15:58:06] <mattz> wrt current ippm reord draft:
[15:58:30] <mattz> current definitions only work with "early" or "late" packets, don't combine. worked really hard to combine in this draft
[15:59:05] <sleinen> Up next: Al Morton http://www.ietf.org/internet-drafts/draft-ietf-ippm-reordering-06.txt
[15:59:18] <mattz> haven't been able to come up with method to combine given current basic reordering defns
[16:00:02] <mattz> [[switch to al morton, update on current draft draft-ietf-ippm-reordering-06.txt]
[16:04:29] <mattz> lots of readers, lots of comments, but no major changes to draft
[16:05:01] --- orange has joined
[16:05:10] --- orange has left
[16:06:11] --- mallman has joined
[16:07:27] <mattz> some comments off-list. david newman (?), henk, ... all clarification
[16:07:36] <mattz> going through laundry list right now
[16:09:13] <mattz> comments based on bartek and przybylski implementation of both this draft, and reordering density (see ippm list).
[16:10:33] <mattz> mentioning comments from david newman on corner cases... now in sect 6
[16:11:32] <mattz> thinks relatively stable (I guess barring reorder density discussion); perhaps last call
[16:11:37] <mattz> no other comments from floor
[16:16:48] <mattz> I did a sketch of the surveyor implementation report. see ippm list message for real details
[16:17:18] <mattz> henk is doing a sketch of the RIPE-TT implementation report
[16:18:01] <mattz> we both implemented everything except type-P-one-way-inverse-percentile :)
[16:18:44] <mattz> we want to collect others... if anyone else has one, see if you can fill out the template
[16:20:03] <mattz> Al morton promises to do in... concentrating on periodic streams and IPDV
[16:20:19] <mattz> s/do in/do one/
[16:20:25] <sleinen> Up next: Emile Stephan on the MIB
[16:20:47] <sleinen> http://www.ietf.org/internet-drafts/draft-ietf-ippm-reporting-mib-06.txt
[16:22:05] <sleinen> Issue: Lack of interest?
[16:24:43] <sleinen> Goes through architecture of the MIB; index allocation etc.
[16:26:41] <mattz> key features of this one (in his view): allows cross-domain measurements by on-the-fly setup (so indecies can have meaning (measurement ID) in 2 domains
[16:31:26] <sleinen> Andy Bierman notes some overlap between this MIB and RMONMIB's TPM MIB
[16:33:51] --- dinakar has joined
[16:41:33] --- dinakar has left
[16:41:40] --- jishac has left
[16:45:25] <sleinen> Now up: Matt Zekauskas on the registry
[16:45:39] <sleinen> http://www.ietf.org/internet-drafts/draft-ietf-ippm-metrics-registry-07.txt
[16:52:13] --- mallman has left: Disconnected
[16:57:22] --- rik wade has left
[17:06:35] --- sleinen has left: Disconnected
[17:14:36] --- mattz has left: Disconnected
[17:18:10] --- dinakar has joined
[17:18:14] --- dinakar has left
[17:31:07] --- mallman has joined
[17:36:01] --- mallman has left
[17:43:09] --- mattz has joined
[18:47:17] --- mattz has left: Replaced by new connection
[18:47:18] --- mattz has joined
[18:47:44] --- mattz has left