[08:00:28] --- iljitsch has become available
[10:55:06] --- jeffa has become available
[10:59:12] --- dmm has become available
[10:59:51] --- suz has become available
[11:02:13] --- Roy Brabson has become available
[11:02:21] --- ggm has become available
[11:02:40] --- javier has become available
[11:02:41] --- sleinen has become available
[11:03:00] --- peetu has become available
[11:03:03] --- FDupont has become available
[11:03:22] <ggm> I'm scribe. I will not be jabbering
[11:03:58] --- dmm has left
[11:04:02] --- momose has become available
[11:04:13] <sleinen> active drafts:
[11:04:19] --- loughney has become available
[11:04:25] <sleinen> draft-ietf-shim6-proto, -hba, -failure-detection, -reach-detect, -arch
[11:04:30] <loughney> Agenda bashing
[11:05:34] --- dumdidum has become available
[11:05:40] <loughney> Assume everyone has read the drafts, this is not a tutorial.
[11:06:06] <loughney> Update: Interim meeting. Minutes sent to the mailing list, design notes sent to the list & folded into the docs.
[11:06:30] <loughney> Pekka Savola - after the meeting, there was a call for consensus. What was the result.
[11:06:32] --- peetu has left: Replaced by new connection
[11:06:48] --- kivinen has become available
[11:06:53] <loughney> Geoff - points were made on this list and folded into the documents. Items without discussion are folded into the disc.
[11:07:06] --- peetu has become available
[11:07:13] <loughney> Review of the WG docs
[11:07:36] <loughney> draft-ietf-shim6-hba presented today draft-ietf-shim6-hba
[11:07:57] <loughney> draft-ietf-shim6-proto will be presented
[11:08:25] --- a2cinoue has become available
[11:08:27] <loughney> draft-ietf-shim6-failure-detection & draft-ietf-shim6-proto-reach-detect presented today
[11:08:39] <loughney> draft-ietf-shim6-arch - work to resume on this.
[11:08:53] --- shep has become available
[11:08:54] --- Roy Brabson has left
[11:09:15] <loughney> expect work to spin up on this.
[11:09:18] <loughney> draft-ietf-shim6-applicability work will resume, see action in the next few months.
[11:09:36] <loughney> draft-ietf-shim6-app-refer - wait until base specs progress
[11:10:01] --- cabo--tzi--org has become available
[11:11:40] --- dumdidum has left: Replaced by new connection
[11:11:49] --- momose has left: Replaced by new connection
[11:12:25] <iljitsch> please use the microphones...
[11:13:05] --- loughney has left: Disconnected
[11:13:38] --- dumdidum has become available
[11:13:43] --- shep has left
[11:16:32] --- andrewdmcgregor@jabber.psg.com has become available
[11:17:19] <iljitsch> ugh
[11:20:22] --- Roy Brabson has become available
[11:20:23] --- shep has become available
[11:22:21] <sleinen> Geoff presents update for draft-bagnulo-shim6-cga-ext-01
[11:23:00] --- oatwillie has become available
[11:24:35] <iljitsch> are the mikes broken?
[11:24:55] <sleinen> No, but people aren't using them (although I'm trying to make them)
[11:25:05] <sleinen> You do understand Geoff though do you?
[11:25:05] --- suz has left
[11:25:28] <iljitsch> yes, it's like there is a miniature Geoff in my computer. :-)
[11:25:39] --- suz has become available
[11:27:05] <sleinen> Up next: Erik Nordmark presenting slides on the protocol update
[11:29:03] --- loughney has become available
[11:29:14] <iljitsch> no url for the slides?
[11:29:34] <loughney> Sorry, had to switch to gprs to get network connectivity ..
[11:30:02] <loughney> cut & paste of older notes ...
[11:30:04] <loughney> draft-bagnulo-shim6-rfc3483-update - passed to ad - waiting for feedback from AD draft- Bob - rfc3483-update - IPv6 wg will fold this into IPv6 Address Selection-bis draft. Margaret - we need to do a revision, will the IPv6 WG be around? Bob - yes, we agreeed on this. Margaret - OK, alright draft-bagnulo-shim5-cga-ext - passed to ad - product of send working, will be handed to another wg draft-huitema-shim6-ingress-filter - no decision yet on it, let's hold off until the base protocols settle down. wg roadmap get the base protocol extension doc to WGLC in decemeber look for further refinements and extensions for 2006 agenda HBA draft only changes from 00 draft are presented locked with cga draft DNS considerations added Updating RFC 3484 - Erik from Marcelo scenario slide
[11:30:49] <iljitsch> thnx
[11:30:55] --- momose has become available
[11:31:06] <iljitsch> there are no proceedings for ietf64...
[11:31:15] <iljitsch> but please let erik continue
[11:31:27] --- FDupont has left: Disconnected
[11:31:28] --- momose has left: Replaced by new connection
[11:31:30] <loughney> There is a new system for real-time procedings, but that's another story ...
[11:31:33] --- andrewdmcgregor@jabber.psg.com has left: Logged out
[11:31:53] <iljitsch> yes, ietf site pretty much never links to anything useful such as the streaming...
[11:32:18] --- dumdidum has left: Replaced by new connection
[11:32:49] --- Roy Brabson has left: Replaced by new connection
[11:32:49] --- jeffa has left: Disconnected
[11:33:15] --- a2cinoue has left: Disconnected
[11:34:39] <loughney> About 3484 - provides an order list and if the app doesn't select the src addr, it picks the 'best' one.
[11:34:39] <loughney> What does 3484 not do?
[11:34:39] <loughney> Bob says it is just the luck of the draw
[11:34:39] <loughney> Erik - yes, - it doesn't know which one works if two address are available & equavilent
[11:34:47] <loughney> guidence to apps - app selects the source addr. recommend that app iterate through the available src addr. The system can return an ordered list of src addr
[11:35:54] <loughney> Selecting the src add
[11:36:13] <loughney> scenario - the app is communicated with a dst but the src addr is unspeficied
[11:36:33] <loughney> signaling chart (can't do ascii art in realtime)
[11:36:48] --- Roy Brabson has become available
[11:37:33] <loughney> Alan Durand - last time we checked, we had problems with timeouts
[11:37:44] <iljitsch> i believe timeout issues are solaris-specific... other OSes do better here
[11:37:50] <loughney> Erik - yes, this would imply modifying tcp
[11:38:00] --- momose has become available
[11:38:09] <loughney> Erik - we don't need to change the standard, but point it out in an appendix.
[11:38:43] <loughney> iljitsch - let me know if you have a mic comment (prepend MIC onto a comment and I'll speak-up)
[11:38:52] <iljitsch> yes I will
[11:39:05] <loughney> Randy Stewart - what is name?
[11:39:12] --- suz has left
[11:39:13] <loughney> Erik - fqdn
[11:39:24] <iljitsch> my take on jabber is that it's also an additional channel usable for commenting on what's going on without interrupting the flow.
[11:39:43] <loughney> - yup
[11:39:54] <loughney> Randy - sctp source address selection is hard.
[11:40:38] <iljitsch> mic comment: before there is a session or association, we don't care yet to which host of the set indicated in the DNS we're talking to.
[11:40:55] <iljitsch> so there isn't really a problem here.
[11:40:57] <iljitsch> (imo)
[11:40:57] <shep> where is his web site?
[11:41:23] <sleinen> www.sctp.org
[11:41:26] <loughney> www.sctp.org
[11:42:02] --- kivinen has left: Disconnected
[11:42:10] --- dumdidum has become available
[11:43:29] <iljitsch> thnx john
[11:43:42] <loughney> Shim6 protocol
[11:43:47] <ggm> I'm taking notes, but have reliable 802.11a. I would willingly switch to jabber if that helps people.
[11:44:09] <loughney> You could take notes on jabber ... and I'll monitor comments and questions
[11:44:35] <ggm> ok.
[11:44:44] <ggm> reviewing unchanged API aspects
[11:45:01] <ggm> host-pair context state unchanged.
[11:45:07] <ggm> whats in the protocol.
[11:45:12] <ggm> packet formats, message exchanges.
[11:45:23] <iljitsch> teardown...
[11:45:28] <ggm> new IP proto number allocated to shim6.
[11:45:33] <ggm> common 16bit header
[11:46:32] <ggm> single bit for control/payload flagging
[11:47:28] <ggm> control messages similar to ICMPv6.
[11:47:59] --- dcrocker has become available
[11:48:17] <ggm> payload MSG. format.
[11:48:25] <ggm> deferred context establishment.
[11:48:45] --- loughney has left: Replaced by new connection
[11:49:05] <ggm> start before negotiate, heuristic used, setup context state. eg 50 packet exchange. then add APIs
[11:49:23] <ggm> API options such as DONTSHIM sock options
[11:49:27] --- loughney has become available
[11:49:39] <ggm> context establishment. the HIP message exchange model
[11:50:18] <ggm> context teardown
[11:50:50] <ggm> uncoordinated. either end can throw away. undesireable when UpperLayer still using. implementation dependent
[11:51:49] --- momose has left: Disconnected
[11:51:51] <ggm> context recovery
[11:52:02] <sleinen> Meta-comment for remote participants: Slides are being uploaded (by Geoff) to https://onsite.ietf.org/public/meeting_materials.cgi?meeting_num=64
[11:52:27] <sleinen> These are already up:
[11:52:30] <sleinen> Agenda: http://onsite.ietf.org/proceedings/05nov/slides/shim6-0.pdf
[11:52:38] <sleinen> HBA: http://onsite.ietf.org/proceedings/05nov/slides/shim6-1.pdf
[11:52:39] <ggm> 3 packet cost to re-establish context
[11:52:49] <sleinen> Proto: http://onsite.ietf.org/proceedings/05nov/slides/shim6-2.pdf
[11:52:51] --- dumdidum has left: Replaced by new connection
[11:52:56] --- dumdidum has become available
[11:52:59] <ggm> context recovery (2) slide
[11:53:18] <ggm> same timing diagram, but now knows peer locaters,
[11:53:26] <ggm> context 'confusion'
[11:54:25] <ggm> what happens when ends remove context un-coordinated. different ULIDs get used for same peer tag
[11:55:48] <ggm> Dino Cisco: why requirement to change IDs?
[11:56:12] <ggm> Erik: not the shim requiring, its an artifact of going outside to DNS. not controlled by SHIM
[11:56:42] <ggm> . but believed to be desireable
[11:57:24] <ggm> Dino understand desire for loadshare, but why not one fixed ID, simplify architecture, if 1:1 mapping
[11:57:42] <ggm> Erik no. SCTP issue. domain doesn't refer to hosts, refers to service on (multiple) hosts.
[11:57:51] * iljitsch is amused
[11:59:33] <ggm> Dino given 32/64 bit ID, assume one, only one host true or false? so have 2 connects from host, can assume if IPs different, different dests (interface IDs) -not up to speed on this.
[11:59:37] <iljitsch> doesn't he mean context id?
[11:59:48] <ggm> (could be)
[12:00:12] --- kurtis has become available
[12:00:13] <ggm> Dino why multiple ID per single host concept? muddies water. puts you into corner cases. trying to keep things simple, deploy quicker..
[12:00:30] <ggm> Erik talk offline, can't get to designated, without putting in DNS, have to do whole set of change not assumed here.
[12:00:46] <ggm> still need recovery mechanism, maybe its simpler.
[12:00:58] <ggm> Dino want to hear compelling reason why the ULID can't stay the same
[12:01:35] <ggm> Erik connect riding on top. can't tell if same box from this layer
[12:01:47] <kurtis> Iljitsch: We should have all slides uploaded
[12:01:59] <ggm> Gih: won't get differnt answer from DNS because its physically different hosts, or MH host with lots of IDs.
[12:02:09] <ggm> tag matching previous contexts, you get lost.
[12:02:32] <oatwillie> er... how does one find the DNS server in the first place?
[12:02:32] <ggm> so slide is about resolving contention. but avoiding new RR in DNS
[12:02:35] <ggm> for host, get same dNS resolution
[12:02:35] <iljitsch> i see hba agenda and proto, I already had proto and hba is already done. :-)
[12:02:51] <ggm> Sending ULP packets
[12:02:54] <ggm> (slide)
[12:02:56] <iljitsch> you may want to add and jari's stuff
[12:02:56] --- kurtis has left: Replaced by new connection
[12:03:06] <iljitsch> ietf64-shim6-rdetect..>
[12:03:11] --- momose has become available
[12:03:27] --- dumdidum has left: Replaced by new connection
[12:03:29] <oatwillie> ggm/ sending packets to find a server/service?
[12:03:38] --- loughney has left: Replaced by new connection
[12:03:38] <ggm> thats the slide title.
[12:03:56] --- dumdidum has become available
[12:04:24] <ggm> if find context, changes src,dst UL to the shim locater pairs.
[12:04:32] <oatwillie> that sounds potentially -BAD-.... (my two year old daughter used to say, "daddy,daddy,daddy,daddy,daddy,daddy,daddy,daddy...."
[12:04:33] <ggm> this is addres-replacement
[12:04:43] <ggm> receiving packets (slide)
[12:05:02] <iljitsch> oatwillie: ?
[12:05:07] <ggm> stax demux on next header values. find shim6, checks payload bit. If set, then control proto, demux and do ctl stuff
[12:05:37] <ggm> for payload, use context tag, to find state. if not found, send R1bis msg. found, replace V6 src, dest with ULID before passing to ULP
[12:06:00] <ggm> Option formats (slide)
[12:06:09] <ggm> Jim: Dino brought stuff up
[12:06:34] <ggm> want to think about.
[12:06:58] --- momose has left: Replaced by new connection
[12:07:14] <ggm> could simplify sending ULP packets case to 2 cases, not 3. just eliminate locater pair == ULID pair.
[12:07:16] --- momose has become available
[12:07:20] <ggm> Erik makes DNS work.
[12:07:25] <ggm> Jim understood. but may be longterm better
[12:07:49] <ggm> could be less complex for small devices.
[12:08:07] <ggm> Pekka Nickander
[12:08:27] <iljitsch> not sure if this needs to be out loud: you don't need to compare both addresses but you can have a rewrite yes/no bit in the context, this is pretty cheap.
[12:08:27] <ggm> as far as I see it, see no obstacles. can be done. had discussion at interim meeting, decided to have it as later item to consider as extension,
[12:08:32] --- Roy Brabson has left: Replaced by new connection
[12:08:42] <iljitsch> we don't want 8 bytes extra for ALL traffic even when there is no failure.
[12:08:44] <ggm> what saying need to establish shim6 state first before can send packets. yes, viable scenario.
[12:08:52] --- Roy Brabson has become available
[12:09:01] <ggm> [flag for somebody to say. maybe somebody else! -ggm]
[12:09:05] <ggm> Erik. in terms of rules
[12:09:21] <iljitsch> NO!
[12:09:33] <ggm> always tagging 8 bytes
[12:09:35] <ggm> [repeat: want it said, be explicit -ggm]
[12:09:51] <ggm> wanna have ULID that are syntactically/functionally disjoint from locaters
[12:10:07] <ggm> if have mech able to do swapping, once in place, can explore what we can do with ULID, ULAs or something else.
[12:10:39] <ggm> Gih: mention in agenda, 2006, contactless shim6 extenstions, how to do things
[12:11:57] --- Roy Brabson has left
[12:12:50] <iljitsch> not kicking in for 3 packets is the whole point...
[12:12:53] <ggm> ggm the 50 number
[12:13:26] <ggm> yes. I now understand. its not a magic number, its an observation, and I understand its not about keepalive in HTTP, even random-spray will get this for 2 endpoints which are in consistent dialogue
[12:13:51] <ggm> it probably means bittorrent won't kick this in :-)
[12:14:06] <ggm> pekka. no interop impact when deciding to make shim
[12:14:15] <ggm> Spencer much discussion on this (out of sequence comment)
[12:14:34] <ggm> forking the context state (slide)
[12:14:51] <ggm> if UL wants to explicitly use different locater pairs, can do so.
[12:15:38] <ggm> capability useful. reachability implications, in other document
[12:15:40] <ggm> Context Tag Re-use (slide)
[12:15:52] <ggm> what if you run out of N bit tags?
[12:15:55] <ggm> when safe to re-use?
[12:16:03] <ggm> Renumbering considerations (slide)
[12:17:08] --- kivinen has become available
[12:17:16] <ggm> (missed slide not discussed)
[12:17:19] <ggm> timers (slide)
[12:17:34] --- Roy Brabson has become available
[12:17:35] <ggm> 4 seconds initial re-tx timer? yes. its good value
[12:17:49] <ggm> requre R1 verifier be usable for minimum time?
[12:17:58] <ggm> yes pick 10sec
[12:18:32] <ggm> size of context tags (slide)
[12:18:34] <ggm> 32 bit. can expand to 47. no packetsize change
[12:18:50] <ggm> worth it, bigger is better
[12:19:07] <ggm> packet injection (slide)
[12:19:17] <ggm> (skipping)
[12:19:20] --- ggm has left
[12:19:33] --- ggm has become available
[12:19:41] <ggm> WHen verify Locaters (slide)
[12:19:53] <ggm> lazy verification. avoid work
[12:20:06] <ggm> but allows on-path attacks until it kicks in
[12:20:19] <ggm> alternative: don't use src locaters to find context
[12:21:46] <ggm> Jari. verify reachability of address in return routability sense
[12:22:21] <ggm> Erik. how to prevent somebody injecting locater packets, but spoofed. whats impact?
[12:22:34] <ggm> Jari. saying if get update, don't verify, then update should not override info.
[12:23:08] <sleinen> btw. we're on slide 27 of http://onsite.ietf.org/proceedings/05nov/slides/shim6-2.pdf
[12:23:39] <sleinen> slide 30
[12:23:52] <ggm> TBDs
[12:23:57] <ggm> (yes, lots of skipping)
[12:24:10] <ggm> consistency with other drafts. need to revserve msg types
[12:24:32] <ggm> Next Steps. fill in TBDs, close open issues. unclosables become FFS. get implementable spec out in a few weeks
[12:26:10] <ggm> Q: how often does this do locater stuff?
[12:26:13] <ggm> Erik: whats the right interzction with sctp?
[12:26:28] <ggm> John. look at tx layer stuff, quick start (c/f Sally Floyd)
[12:26:44] <ggm> Erik: good to start collecting what are types of things to cover in general API
[12:27:03] <ggm> Spencer like work. but tx work could be done, should be done, hasn;t been done yet.
[12:27:38] --- nm has become available
[12:28:20] <ggm> Kurtis. API doc, do next year. volunteer doc author (Pekka) take input to ML. running short of time
[12:28:27] <ggm> Erik can shift fixed 4 to random between 3-5
[12:28:35] --- dcrocker has left
[12:28:36] <ggm> Spenser. good to get somebody from tx to review
[12:28:41] <ggm> avoid fixed timers
[12:29:14] <iljitsch> timer issues are also important with regard to implementation...
[12:29:18] <ggm> Dino: how rp work when MultiHop not over TCP with packetloss
[12:29:48] <ggm> Erik: don't understand. from RP looks like multi-add sending, routing doesn't care. take offline
[12:30:26] <ggm> Jari Arkko on reachability
[12:30:34] <ggm> 02 draft, in good shape
[12:31:24] <ggm> reviewing protocol history
[12:31:44] <ggm> having problems describing a state machine,
[12:32:06] <ggm> Erik suggests avoid state machine, but protocol has holes
[12:32:25] <iljitsch> what was that?
[12:32:28] <ggm> IETF64, using Eriks model refinee details, complete (?) but could be mistaken
[12:32:37] <ggm> glasses next door on table I think. water table collapse? sounds like it
[12:32:45] <ggm> (so it made the mike!)
[12:32:57] <ggm> diagram.. aaargh!
[12:33:09] <ggm> Scope of SHIM6 work (slide)
[12:33:42] <ggm> design decisions (slide)
[12:34:01] <ggm> avoid config, protocols, do not re-invent DHCP, beleive what ND tells us
[12:34:15] <ggm> own addresses learned locally, ppeer addresses communicated
[12:34:28] <ggm> shim6 only works as failover
[12:34:35] <ggm> FBD is chosen for simplicity
[12:34:57] <ggm> Other design goals (slide)
[12:35:03] <ggm> avoid packets if UL idle.
[12:35:07] <ggm> piggyback
[12:35:23] <ggm> handle unidirectional failures.
[12:35:33] <ggm> provide return route verification at the same time
[12:35:43] <ggm> separateable component, can reiuse in other contexts
[12:35:58] <ggm> (personal goal)
[12:36:25] <ggm> Reachability vs Exploration (slide)
[12:36:34] <ggm> c/f Iljitsch
[12:36:59] --- Roy Brabson has left
[12:37:31] <ggm> Proto case 1. idle (slide)
[12:37:37] <ggm> Proto case 2: bidi (slide)
[12:37:42] <ggm> if bidi, seen, then fine
[12:37:47] <ggm> Proto 3, unidir (slide)
[12:38:01] <ggm> need keepalive (id=10)
[12:38:09] <ggm> return pkt
[12:38:51] <ggm> Proto case 4 (slide)
[12:38:58] <ggm> unidir failure
[12:39:07] <ggm> packetflow back to log failure
[12:39:16] <ggm> (on timer fire)
[12:39:20] <ggm> 'event' message
[12:40:35] <ggm> Case 4 contd (slide)
[12:40:50] <ggm> showing recovery state
[12:41:32] <ggm> Behaviour in the general case (slide)
[12:41:51] <ggm> had state machine in -02, broken., too complicated
[12:42:18] --- loughney has become available
[12:42:21] <ggm> get info in indirect manner. cannot use request/response
[12:42:32] <ggm> how to stop exploration when both happy
[12:42:38] <ggm> anything can happen at any time
[12:42:58] <ggm> (slide of state machine)
[12:43:06] <ggm> Simplified descr (slide)
[12:44:16] <sleinen> [We're on slide 19 of http://onsite.ietf.org/proceedings/05nov/slides/shim6-4.pdf]
[12:44:32] <ggm> Exploration part (slide)
[12:44:42] <ggm> discussing the simplified model.
[12:46:31] <ggm> Simplified description (slide)
[12:46:33] <sleinen> [slide #21]
[12:46:53] <ggm> Issues to think about (slide)
[12:46:53] <sleinen> [slide #22]
[12:47:16] <ggm> how to stop exploration (slide)
[12:47:32] <ggm> as described, continues forever. needs fixing (!)
[12:48:12] <ggm> key Q. when to use, not how represented
[12:48:34] <ggm> skpped to keepalive and exploration integrated (slide)
[12:48:43] <ggm> recommended to keep separate
[12:50:08] <ggm> processes per context or per host? (slide)
[12:50:21] <ggm> per host approach (Iljitsch)
[12:51:24] <ggm> Gih: WG hat off. appoach here is base spec, conservative, but works. equivalences of sets of ULID, other end all the same too, making life much more complicated, understanding this. per-context is conservative approach. do per context at this point, on that basis
[12:51:40] --- lixia has become available
[12:51:57] <iljitsch> mike comment: I believe (but not sure) state diagram will be simpler by doing some stuff per-host rather than per-context
[12:52:03] <ggm> Timing issues, other layers (slide)
[12:52:45] <ggm> timescale, interactions with other protocols. eg TCP tcp retx. TE mechanisms.
[12:53:26] <ggm> probably should set limits to not act faster than tcp re-tx, later than TCP give-up
[12:53:38] <ggm> Tim Shepard. thinking we have deeper problem here
[12:53:52] <ggm> reason pair of locaters may appear to fail is congestion on one connection off site
[12:54:21] <ggm> imagining thousands of hosts doing this, all do 'time to do shim6 failover' move to other thing.. dynamics could be complicated
[12:54:28] <ggm> who is responsible for getting this right. in scope?
[12:54:45] --- loughney has left: Disconnected
[12:54:46] <ggm> Gih: touching on TE, load balance. not directly addressing. but next year comes back in scope
[12:54:49] --- loughney has become available
[12:55:18] <ggm> Loughney look at something conservative
[12:55:48] <ggm> avoid webserver doing a shim move on all connects. some use of a bit of randomization to do this.
[12:55:50] <ggm> Jari yes
[12:55:56] <ggm> add this at first, without good metrics
[12:56:24] <ggm> Spenser: IAB note, not thought about it?
[12:56:45] <ggm> Gih: work earlier deciding direction, basic conservative approach, IESG would have looked at it, be informed
[12:56:48] <ggm> Pekka is taking notes
[12:56:56] --- FDupont has become available
[12:57:02] --- obione has become available
[12:57:33] --- josoinin has become available
[12:57:57] --- Roy Brabson has become available
[12:58:08] <ggm> ggm clarifies 'conservative' means 'punt for now, keep WORK simple'
[12:58:20] <ggm> Issues to merge (slide)
[12:58:34] <ggm> newest version has some new things.
[12:58:50] <ggm> Ilitsch has mechanism to allow separation of real data packets, vs TCP acks.
[12:59:13] <ggm> Jari wrong person to talk, don't understand how works. [Gih has presentation to follow from Ilitsch]
[12:59:20] <ggm> bunch of details on heuristics, preferneces.
[12:59:26] <iljitsch> the data / non-data stuff isn't discussed in the presentation, though.
[13:00:15] <ggm> [hey, life is full of surprises -:-) -ggm]
[13:00:20] <ggm> Next Steps (slide)
[13:00:42] <ggm> update draft. review, WGLC. want verification (formal?) make sure not forgetting states
[13:01:19] <ggm> issues ..
[13:01:42] <ggm> Andrew Mcready. issues with congestion, timers
[13:02:04] <ggm> all timers changing paths, should have some randomness. otherwise its synchronized death (VJ)
[13:02:14] <ggm> (he called it a firefly oscillator)
[13:02:28] <iljitsch> 01
[13:02:43] <ggm> Ilitsch has done 01 draft of reachability John will present
[13:02:50] <ggm> [clarified -ggm]
[13:03:02] <ggm> reach-detect
[13:03:07] <ggm> John Loughney presents
[13:03:16] <ggm> (un)reachability (slide)
[13:03:32] <ggm> preferences (slide)
[13:03:59] <ggm> strong/weak values. strong (p1) hgiher is better. weak (p2) load balancing weight.
[13:04:06] --- dthaler has become available
[13:04:08] <ggm> unidirectional reachability (slide)
[13:04:14] <sleinen> [This is http://onsite.ietf.org/proceedings/05nov/slides/shim6-5.pdf (as you might have guessed)]
[13:04:17] <ggm> uncommon for normal failures
[13:04:56] <ggm> keepalives (slide)
[13:04:58] <ggm> [I need to go out for a mo, can somebody else do this? -ggm]
[13:05:13] <iljitsch> no we don't test when there is no traffic...
[13:05:38] <ggm> [ok. unavoidably offline for 10 min -ggm]
[13:05:43] <iljitsch> this is the same as what Jari talked about , btw.
[13:07:43] <dthaler> dino: avg web session is 17 packets, assume no keepalives during that time
[13:07:56] <iljitsch> no receiver NEEDS to see data or keepalives every 10 seconds so no backoff
[13:08:04] <dthaler> dino: is there an exponential backoff on the timer?
[13:08:38] <iljitsch> (is anyone relaying?)
[13:08:57] <sleinen> Iljitsch: do you want me to ("mic")?
[13:09:09] <iljitsch> yes my comment on the backoff
[13:10:37] <dthaler> jari: quick check is a heuristic, possible to design a protocol which does exploration as part of quick check
[13:11:27] <dthaler> now on slide 8 (fields)
[13:11:52] <sleinen> [slide #10]
[13:12:43] <sleinen> [#11]
[13:12:44] <ggm> [bad news: I just got late checkout declined (!) -they have 300 new guests arriving today. I have to go pack -ggm]
[13:13:17] <iljitsch> 10 minutes, not bad!
[13:13:28] --- peetu has left: Logged out
[13:13:39] <iljitsch> I want to do another clarification and then we need to do the big merge.
[13:13:45] <iljitsch> I need feedback though.
[13:13:56] <iljitsch> (relay please)
[13:14:29] <dthaler> geoff: per host vs per context... propose WG just does per context for now and leaves per host for later
[13:14:38] <ggm> next: misc shim6 steps
[13:15:00] <sleinen> http://onsite.ietf.org/proceedings/05nov/slides/shim6-6.pdf
[13:15:00] <dthaler> (loughney relayed)
[13:16:40] <ggm> geoff has notes, slides. site exit path selection
[13:16:45] <ggm> ingress filtering
[13:17:02] <ggm> can present wrong packets. enable host to figure out what best way of getting packet out of site
[13:17:22] <ggm> assertion in draft, some form of SAD source address dependent (is acronym deliberate) needed
[13:17:52] <ggm> [ggm offline. sorry]
[13:18:14] --- ggm has left
[13:19:31] <sleinen> Dino comments: was it considered to put (S,G) routing into the IGPs(?)?
[13:19:32] <dthaler> dino: was it an alternative considered to put (S,D) routing in the IGP?
[13:19:51] <sleinen> (S,D) of course, Dave's right
[13:22:52] <dthaler> nordmark: can get rid of no-context error and do R1bis instead since simpler to describe those rules
[13:22:56] --- loughney has left
[13:23:41] <dthaler> we are done
[13:23:51] --- dthaler has left
[13:23:59] --- javier has left: Disconnected
[13:23:59] --- FDupont has left: Disconnected
[13:24:07] <iljitsch> thanks for relaying and presenting, guys
[13:24:12] --- Roy Brabson has left
[13:24:35] --- nm has left
[13:24:38] --- dumdidum has left
[13:24:52] --- shep has left
[13:25:19] --- iljitsch has left
[13:25:42] --- sleinen has left
[13:26:42] --- obione has left: Disconnected
[13:27:17] --- oatwillie has left
[13:27:19] --- Yoshifumi Atarashi has become available
[13:27:23] --- cabo--tzi--org has left
[13:29:39] --- kivinen has left
[13:29:53] --- Yoshifumi Atarashi has left
[13:30:08] --- josoinin has left
[13:39:52] --- momose has left
[13:47:25] --- lixia has left
[16:25:36] --- dcrocker has become available
[18:35:00] --- dcrocker has left