IETF
mif@jabber.ietf.org
Tuesday, July 21, 2015< ^ >
Dave Thaler has set the subject to: IETF-90 Toronto meeting in progress
Room Configuration
Room Occupants

GMT+0
[13:17:25] Meetecho joins the room
[13:22:56] Dave Thaler joins the room
[13:23:08] Dave Thaler has set the subject to: Multiple Interfaces WG
[13:23:39] <Dave Thaler> chairs slides: https://www.ietf.org/proceedings/93/slides/slides-93-mif-1.pdf
[13:23:50] <Dave Thaler> agenda bashing
[13:27:39] <Dave Thaler> Erik Kline's slides on API: https://www.ietf.org/proceedings/93/slides/slides-93-mif-2.pdf
[13:35:23] Dan Wing joins the room
[13:36:01] Jonathan Lennox joins the room
[13:36:16] <Dave Thaler> slide 5: new things to define
[13:39:59] <Dave Thaler> Ian Farrer at mic
[13:41:33] <Dave Thaler> Steven Barth at mic
[13:42:06] <Dave Thaler> Margaret Wasserman
[13:44:30] <Dave Thaler> Suresh Krishnan at mic
[13:44:35] <Dave Thaler> Steven Barth
[13:44:50] Ted Lemon joins the room
[13:47:12] <Dave Thaler> Danny Moses at mic
[13:48:03] <Dave Thaler> Margaret Wasserman/Cullen
[13:50:53] <Dave Thaler> Mark Townsley
[13:51:38] <Dave Thaler> Lorenzo Colitti
[13:52:38] <Dave Thaler> Mark & Margaret at mic
[13:54:25] <Dave Thaler> Lorenzo
[13:55:43] <Dave Thaler> Steven Barth
[13:57:36] Jonathan Lennox leaves the room
[14:00:49] <Dave Thaler> Lorenzo
[14:02:04] <Dave Thaler> Markus Stenberg
[14:05:26] <Dave Thaler> Mark Townsley
[14:06:16] <Dave Thaler> (30 minutes so far just on this one slide...)
[14:08:38] Brian Haberman joins the room
[14:11:21] <Dave Thaler> Suresh
[14:13:28] <Dave Thaler> slide 6: some sockets API considerations
[14:14:59] <Dave Thaler> 7: sockets API
[14:17:34] <Dave Thaler> Dave Thaler at mic
[14:17:37] <Dave Thaler> now Mark Townsley
[14:19:12] <Dave Thaler> 8: sockets API (socket/bind)
[14:19:37] <Dave Thaler> 9: listen/accept/connect
[14:20:31] <Dave Thaler> 10: sendmsg/recvmsg
[14:20:55] <Dave Thaler> 11: DNS resolution
[14:21:24] Joe Crowe joins the room
[14:22:02] <Dave Thaler> 12: next steps
[14:25:38] <Ted Lemon> that can be done through ISE.
[14:25:42] <Ted Lemon> which is a really good idea.
[14:27:56] <Ted Lemon> I think the sockets API is a bad API for MIF.
[14:28:20] <Ted Lemon> I don't think it's actually possible to do MIF with the sockets API.
[14:34:33] Keith Moore joins the room
[14:35:53] Joe Crowe leaves the room
[14:36:07] <Dave Thaler> need a mic relay ted?
[14:36:26] Dan Wing leaves the room
[14:38:19] <Ted Lemon> hummmmm
[14:38:38] <Ted Lemon> I think that we should feed the two processes into each other.
[14:38:49] <Ted Lemon> So yes to the current question.
[14:38:54] <Dave Thaler> which two?  (please use "mic:" prefix if oyu want relay)
[14:39:19] <Ted Lemon> mic: I think we should do both abstract and concrete in parallel so that each effort can infom the other.
[14:39:27] <Ted Lemon> (thanks)
[14:39:29] <Keith Moore> mic: whatever else you do, please don't delay development of the sockets api
[14:39:47] <Dave Thaler> mic ack x2
[14:39:55] <Keith Moore> (thanks)
[14:40:22] <Ted Lemon> mic: not a sockets api
[14:40:27] <Ted Lemon> mic: a concrete api
[14:40:55] <Ted Lemon> right
[14:41:14] <Ted Lemon> mic: we can just publish it through the ISE after it's done.
[14:41:18] <Dave Thaler> Terry manderson at mic
[14:41:28] <Dave Thaler> (AD hat)
[14:41:32] <Keith Moore> mic: I agree that a concrete API is necessary, but I also think a sockets API is necessary
[14:42:07] <Ted Lemon> Keith, no need for mic, but what is the problem you are trying to solve?   I think cramming this into the sockets API is likely not to be very useful, but it sounds like you think it would be useful.
[14:42:08] <Dave Thaler> can you elaborate on the diff kewith?
[14:42:46] <Keith Moore> I basically think that if you don't have a sockets API you don't have any basis for writing portable code that uses it.
[14:43:08] <Ted Lemon> so isn't what you are asking for a portable API?
[14:43:20] <Keith Moore> yes, but it has to use the sockets interface.
[14:43:21] <Ted Lemon> The reason I'm holding your feet to the fire is that sockets have a lot of problems.
[14:43:24] <Ted Lemon> Why?
[14:43:43] <Ted Lemon> What is it about the sockets API that you are afraid of losing?
[14:43:44] <Keith Moore> because it's the closest thing that exists to a universal api
[14:44:05] <Keith Moore> though I certainly agree that there are shortcomings with the sockets api
[14:44:09] <Ted Lemon> E.g., is it because you want the connection to be bound to a single descrptor that you can do sendmsg() on?
[14:44:20] <Ted Lemon> or is it that you want socket(), bind() and accept()?
[14:44:29] <Ted Lemon> or setsockopt, or something? :)
[14:45:36] <Ted Lemon> mic: I would like to be able to write a concrete API that is compatible with the abstract API.
[14:45:38] <Dave Thaler> FYI: https://en.wikipedia.org/wiki/Austin_Group
[14:45:46] <Keith Moore> It needs to be compatible with file descriptors and support read/write/sendmsg/recvmsg/shutdown/close.  if you have to make different calls to setup a connection, that uses this api, that doesn't bother me too much.
[14:45:58] <Ted Lemon> keith:okay, then I think we actually agree.
[14:46:14] <Keith Moore> ted: sounds good
[14:46:19] <Ted Lemon> hm, that was an unintended emoticon!
[14:46:41] <Dave Thaler> Organisation     Representative
ISO/IEC JTC 1/SC 22     Nick Stoughton
IEEE PASC     Don Cragun
The Open Group     Martin Rehak
[14:47:54] <Dave Thaler> Homenet use cases presentation: https://www.ietf.org/proceedings/93/slides/slides-93-mif-3.pdf
[14:48:16] <Dave Thaler> slide 2: PvD in Homenet
[14:49:21] <Brian Haberman> Yep, lwig
[14:49:37] Brian Haberman leaves the room
[14:50:01] <Dave Thaler> 3: Conveying PvD info
[14:50:33] <Dave Thaler> 4: Next steps
[14:51:27] <Dave Thaler> Teco Boot at mic
[14:52:43] <Dave Thaler> Mark Townsley responding
[14:53:42] <Dave Thaler> Ray Bellis & Mark Townsley at mic
[14:55:32] <Dave Thaler> Markus Stenberg
[14:56:29] <Dave Thaler> Steven Barth
[14:57:49] <Ted Lemon> mic: I think we want homenet routers to be mif-capable.
[14:58:11] <Ted Lemon> mic: perhaps that is the wrong thing, but I don't think that is obvious.
[14:58:21] <Dave Thaler> Suresh Krishnan
[14:58:23] <Dave Thaler> mic ack
[14:58:45] <Dave Thaler> (Ted is after steven after suresh)
[14:58:54] <Dave Thaler> steven now
[14:59:30] <Ted Lemon> Huh, looks like the t-shirt this time was pretty good.
[15:00:51] <Dave Thaler> markus jumping queue...?
[15:00:56] <Keith Moore> (maybe we need cool t-shirts for remote attendees)
[15:01:24] <Ted Lemon> My IETF colleagues went to Prague and all I got was this lousy T-shirt?
[15:01:37] <Keith Moore> agree with you Ted, about wanting homenet routers to be mif-capable
[15:01:45] <Ted Lemon> mic: I mean that we should consider the question.
[15:01:55] <Ted Lemon> mic: my personal opinion is that we want them to be capable.
[15:02:03] <Ted Lemon> mic: but I am not sure that this is true.
[15:02:22] <Ted Lemon> mic: the reason I want it is that it's pretty clear what we need based on the mpvd arch.
[15:02:38] <Dave Thaler> ack
[15:03:02] <Ted Lemon> this should work with automatic pvds as well as provider-provided pvds.
[15:04:12] <Ted Lemon> mic: homenet basically doesn't work well without MIF
[15:05:29] <Ted Lemon> mic: the thing that routers need to do is to come up with pvd ids for the providers they are connected to.
[15:05:36] <Dave Thaler> ack
[15:05:48] <Ted Lemon> mic: because we are trying to make homenets work with status quo ISPs.
[15:06:02] <Ted Lemon> thanks for all the mic activity, Dave.
[15:06:08] <Dave Thaler> no prob
[15:06:28] <Ted Lemon> mic: so the answer to that is that there is more than just DHCP options.
[15:06:33] <Dave Thaler> suresh at mic
[15:06:49] <Ted Lemon> mic: the routers may not get any information, but still need to announce different PVDs for each provider.
[15:07:18] <Dave Thaler> still in line...
[15:07:24] <Ted Lemon> I can see!
[15:08:34] <Ted Lemon> *thud*
[15:08:46] <Dave Thaler> Mark Townley from floor "there's work to do :)"
[15:09:06] <Dave Thaler> Terry Manderson (AD) at mic
[15:10:22] <Keith Moore> I don't get the point.
[15:10:35] <Ted Lemon> The mailing list is moribund.
[15:10:38] <Ted Lemon> It's a valid criticism.
[15:11:17] <Ted Lemon> FWIW, this was a very productive session, but I would have liked to have heard something about the API work earlier.
[15:11:20] <Dave Thaler> meeting adjounred
[15:11:39] <Dave Thaler> jabber scribe signing off :)
[15:11:45] <Ted Lemon> Thanks dave!
[15:12:37] Ted Lemon leaves the room
[15:13:23] Keith Moore leaves the room
[15:20:13] Meetecho leaves the room
[15:23:39] Dan Wing joins the room
[15:25:07] Dan Wing leaves the room
[15:26:07] Dave Thaler leaves the room
[15:34:45] Jonathan Lennox joins the room
[15:37:01] Dan Wing joins the room
[15:37:16] Dan Wing leaves the room
[15:41:21] Jonathan Lennox leaves the room
[15:55:46] Dave Thaler joins the room
[15:55:54] Dave Thaler leaves the room