[18:48:16] --- Tom Phelan has joined
[18:50:57] --- admcd has joined
[18:54:11] <admcd> most slides appearing at: http://www.tschofenig.com/nsis/IETF60
[18:55:24] <admcd> john loughney clarification to jabber minutes yesterday: there are people working on designing;/developing implementations
[18:55:45] <admcd> robert: we're open to discussions if people want to come talk to us
[18:56:01] <admcd> nslp for qos signalling - sven van den bosch presenting
[18:56:41] <admcd> two versions since seoul - -03 for interim, now got -04
[18:57:28] <admcd> receiver initiated reservation mechanisms updated
[18:57:44] <admcd> session binding now split from fate sharing
[18:58:04] <admcd> flag sense reversed (now all positive, rather than NO_xyz)
[18:58:12] --- loughney has joined
[18:58:24] <admcd> fate sharing does not explicitly support fate sharing - can do at endpoints
[18:58:35] <admcd> refresh reduction updates
[18:59:11] <admcd> priority issues - message priority (transfer attributes over gimps api), reservation prority (qspec issue)
[18:59:35] <admcd> clarifications on response-request (RII)
[18:59:37] --- mankin has joined
[18:59:49] <admcd> object formats updated - following format used by other drafts
[18:59:57] --- Tom Phelan has left: Disconnected.
[19:00:07] <admcd> closed various open issues as a result of these updates
[19:00:49] <mankin> The AD is detained in the hall and may not make it in
[19:01:01] <admcd> other changes to -04
[19:01:44] --- danwing has joined
[19:02:07] --- falk has joined
[19:02:10] <admcd> added some initial work on an ERROR_SPEC object
[19:03:00] <admcd> first byte indicates a 'severity' - informational, success, protocol error, etc
[19:03:58] <admcd> mailing list question on adding a "meaning" object - issue still open
[19:04:20] <admcd> question of whether to use flags, "meaning" object, or separate message types
[19:04:36] --- Tom Phelan has joined
[19:04:40] <admcd> open issues
[19:04:57] <admcd> mainly authorization - see later discussion in this meeting
[19:05:34] <admcd> error codes - is basic structure right? also need to include in message/object processing text in draft
[19:06:05] <admcd> ruediger geib: two major points
[19:06:22] <admcd> default behaviour of reserve is to not be acked
[19:06:45] <admcd> default should be to always be acked
[19:07:08] <admcd> you also now use QUERY as a 'reserve'
[19:08:26] <admcd> sven: does not by default not send response - see applications where this may be appropriate
[19:08:59] <admcd> sven: second point - QUERY causes reverse-path state installation at NTLP, not install state at NSLP
[19:09:55] <admcd> bob braden: comment on error codes
[19:10:13] <admcd> from rsvp: turns out to bigger issue than looks
[19:10:28] <admcd> start work on iana considerations now - this bit us later in rsvp
[19:11:05] <admcd> next up: qspec template (presented by attila bader)
[19:11:37] --- falk has left: Disconnected
[19:13:40] <admcd> qspec template defines generic/optional parameters
[19:15:08] <admcd> qos description = <qos desired> <qos available> <qos reserved> <minimum qos>
[19:15:26] --- falk has joined
[19:15:30] <admcd> bob braden: your use of bnf suggests an ordering
[19:15:56] <admcd> attila: no ordering is required at present
[19:16:41] <admcd> open issues
[19:17:25] <admcd> need to clarify relationships between various parts of qsm/qos-nslp processing
[19:17:54] <admcd> reservation priority
[19:18:14] <admcd> do we need traffic specification or qos desired?
[19:18:24] <admcd> propose as wg draft
[19:18:38] <admcd> jl: at interim we thought it would be useful
[19:18:52] <admcd> jl: please comment on whether this is good/bad idea on mailing list
[19:19:32] <admcd> rmd - resource management in diffserv (attila presenting)
[19:19:53] <admcd> (request for mic to be moved up, or attila's mouth to be moved down) (laughter ensues)
[19:20:26] <admcd> many different diffserv qos models are possible
[19:20:44] <admcd> rmd is an example of a diffserv signaling model
[19:21:25] <admcd> edge to edge signalling in diffserv domain
[19:21:37] <admcd> can be combined with end-to-end signalling
[19:23:26] <admcd> propose rmd draft for wg
[19:23:33] <admcd> jl: discussed at interim
[19:23:53] <admcd> jl: but, do we want general diffserv model, or rmd?
[19:24:13] <admcd> rudi: currently to thin to make wg document
[19:24:24] <admcd> jl: authors apparently have plans to flesh out
[19:24:33] <admcd> jl: should be that version that is adopted
[19:24:47] <admcd> brian carpenter: exchair diffserv, wasn't at interim
[19:25:12] <admcd> bc: I need to read documents. rings bells. per flow signalling.
[19:25:22] <admcd> tom phelan: confused
[19:25:37] <admcd> tp: jl you made it sound like an either-or between general/specific diffserv models
[19:25:58] <admcd> jl: not necesarily either-or, should be discussed
[19:26:24] <admcd> bb: maybe we should have objective to simplify the whole picture to a unified model
[19:26:31] <admcd> jl: that sounds pretty difficult
[19:26:56] <admcd> ab: want to define concrete protocol
[19:28:02] <admcd> extended qos quth - hannes tschofenig presenting
[19:28:36] <admcd> currently have two party, token based three party and generic three party authz approaches
[19:29:06] <admcd> this draft looks at third of those using chall/resp or generic model based on eap
[19:30:49] <admcd> for three party model
[19:30:59] <admcd> end host must take actively take part
[19:31:12] <admcd> true auth between end host and aaa server
[19:31:34] <admcd> difference between eap and c/r is mainly in terms of flexibility
[19:32:13] <admcd> for c/r mechanism have to pick a particular c/r mechanism
[19:33:12] --- falk has left: Disconnected
[19:33:17] <admcd> msg flow is similar for eap - example here based on 3gpp aka
[19:33:56] <admcd> there are some technical issues with both c/r and eap
[19:34:08] <admcd> binding nslp and lower layer security mechs together
[19:34:47] <admcd> combination of ntlp unilateral auth and client layer mechanisms
[19:35:03] <admcd> next step: decide which way to go
[19:35:34] <admcd> jl: i suggest you read the draft, it isn't very long and is fairly clear
[19:36:07] <admcd> jl: main open issue in this area. would like wg consensus before next meeting.
[19:36:14] <admcd> next up:
[19:36:25] <admcd> nslp for accounting config (still hannes presenting)
[19:36:37] <admcd> jl: would suggest maybe metering instead of accounting
[19:36:54] <admcd> (murmurs from room)
[19:37:10] <admcd> this is another possible application for nsis
[19:37:20] --- mankin has left
[19:37:29] <admcd> dynamic config of accounting entities rather than hardwiring them
[19:38:26] <admcd> scenario 1: charging
[19:38:32] <admcd> related to qos
[19:39:01] <admcd> selectively accounting for specific flows from entities along data path
[19:39:21] <admcd> scenario 2: monitoring
[19:39:26] <admcd> (ref ipfix)
[19:39:55] <admcd> accounting data for monitoring, e.g. whether sla is fulfilled
[19:40:28] <admcd> has anyone read this draft?
[19:40:34] <admcd> about five people raised hands
[19:40:45] <admcd> jl: read and comment on mailing list
[19:41:08] <admcd> jl: remain as individual draft for now, want to concentrate on current work first
[19:41:36] <admcd> tp: second question on slide was interesting - how to add new nslps
[19:41:46] <admcd> jl: want to concentrate on current work first
[19:41:57] <admcd> jl: maybe give guidelines for adding nslps later
[19:42:13] <admcd> diameter qos application (peter mccann presenting)
[19:42:16] <admcd> discussed in aaa
[19:43:12] <admcd> why needed?
[19:43:30] <admcd> authentication, authorization, accounting for qos resvs and interface to application servers
[19:44:43] <admcd> allow application servers to dynamically authorize/de-auth flows
[19:46:21] <admcd> without this every domain deploying qos needs an application server
[19:46:22] <admcd> may mean sip is only way to ask for qos
[19:46:30] <admcd> problems:
[19:46:37] <admcd> couples application to signalling path, etc
[19:47:11] <admcd> with diameter qos can mediate between things on path, application servers, subscriber database
[19:47:43] <admcd> advantages:
[19:47:53] <admcd> uniform i/f for aaa of qos
[19:48:08] <admcd> application servers can be located independently from data path
[19:49:00] <admcd> work in progress:
[19:49:12] <admcd> representation of app flows (pkt filters, qos description, etc)
[19:49:21] <admcd> moving forward
[19:49:32] <admcd> progress as individual draft
[19:50:25] <admcd> unknown: don't see why this is needed, can be solved through resource management
[19:51:52] <admcd> jl: we're done
[19:52:04] <admcd> jl: before washington want to make progress on main drafts
[19:52:17] <admcd> jl: finalise main ntlp, nslp drafts
[19:52:22] --- Tom Phelan has left
[19:52:33] <admcd> jl: please read, comment on drafts that have been proposed as wg items
[19:52:46] <admcd> jl: see you in washington in november
[19:56:47] --- loughney has left
[20:01:15] --- danwing has left: Disconnected
[20:05:44] --- admcd has left