Tuesday, 6 November 2012< ^ >
Room Configuration

[17:53:15] Florin Coras joins the room
[18:00:15] Luigi Iannone joins the room
[18:01:42] damien.saucez joins the room
[18:02:27] <damien.saucez> delayed by 5min to let people arriving
[18:03:16] <damien.saucez> blue sheets
[18:04:14] Eliot Lear joins the room
[18:04:24] Wassim M. Haddad joins the room
[18:04:39] <damien.saucez> chair welcomes
[18:05:12] <Eliot Lear> you DO have remote participants!
[18:06:02] <damien.saucez> is streaming working well?
[18:06:07] <Eliot Lear> doing just fine
[18:06:16] <damien.saucez> 7 drafts in the RFC queue
[18:06:26] <damien.saucez> currently wedged on UDP checksum
[18:07:18] <damien.saucez> change of status so WG last call again
[18:08:03] <damien.saucez> agenda
[18:08:30] <damien.saucez> brian: waiting for document
[18:08:38] <damien.saucez> chair: will follow that up
[18:08:47] <damien.saucez> back to agenda bashing
[18:09:04] <damien.saucez> 5 items today
[18:09:15] <Eliot Lear> hands are up
[18:09:20] <damien.saucez> who read the documents?
[18:09:23] <damien.saucez> about 12
[18:09:36] <Eliot Lear> i have read both intro and arch docs
[18:09:39] <damien.saucez> architecture: 12 again
[18:09:49] <damien.saucez> keep hands if want to discuss this today
[18:09:50] <damien.saucez> yes
[18:09:56] <Eliot Lear> is noel in the room?
[18:09:57] <Florin Coras> me too
[18:10:12] <damien.saucez> no special request from attendees
[18:10:16] <damien.saucez> dino on the stage
[18:10:41] <damien.saucez> joke about chocolate cookies
[18:10:57] <Eliot Lear> would someone please pass the cookie over the jabber connection?
[18:11:19] <damien.saucez> chair says No :-
[18:11:48] <damien.saucez> lcaf: became wg draft
[18:12:07] <damien.saucez> initial draft april 2010
[18:12:17] <damien.saucez> accepted to have an AFI from Iana
[18:12:41] <damien.saucez> (discussion about a typo)
[18:12:53] <damien.saucez> actually no! AFI received before the draft
[18:13:21] <damien.saucez> idea: possibility to extend for new use cases
[18:13:25] <damien.saucez> slide 5
[18:14:03] <damien.saucez> slide 6
[18:14:45] <damien.saucez> nat traversal LCAF will be discussed later
[18:14:47] <damien.saucez> slide 7
[18:15:31] <damien.saucez> after experiments, decided to add altitude to geo-coordinates
[18:15:53] <damien.saucez> ELP: explicit locator
[18:16:22] <damien.saucez> DDT-SEC has been presented a couple of IETF ago
[18:16:23] <damien.saucez> slide 8
[18:17:18] <damien.saucez> not a lot of details about the types in the draft, for that go to individual drafts for each type
[18:17:56] <damien.saucez> slide 9
[18:20:37] <damien.saucez> to support a new type, need to change the mib, so better to use lcaf (more flexible)
[18:21:17] <Eliot Lear> there's no benefit of using inetAddress if every network management implementation has to change ANYWAY. might as well use LCAF
[18:21:26] <damien.saucez> has to improve text for security
[18:21:36] <damien.saucez> Luigi on stage
[18:21:57] <damien.saucez> slide 1
[18:22:15] <damien.saucez> quick update on 2 update of the draft
[18:22:17] <damien.saucez> slide 2
[18:22:49] <damien.saucez> addition of a new attack, actually combination of two attacks already in the draft
[18:22:54] <damien.saucez> slide 3
[18:23:18] <damien.saucez> example of the attack
[18:24:33] <damien.saucez> LISP sec can be used to protect against this
[18:24:36] <damien.saucez> slide 4
[18:24:49] <damien.saucez> editorial polishing
[18:25:34] <damien.saucez> removed ref to RFC2119 as never use these terms
[18:26:03] <damien.saucez> future plans removed as believed the draft is at its end
[18:26:05] <damien.saucez> slide 5
[18:26:35] <damien.saucez> question: last call?
[18:26:52] <damien.saucez> darrel to the mic
[18:27:10] <damien.saucez> text needs work
[18:28:05] <damien.saucez> darel: ok for last call as the change, but document should discuss more the difficulty of leading the attacks
[18:28:26] <damien.saucez> darrel (e.g., is miss configuration enough or really need effort )
[18:28:39] <damien.saucez> luigi: yes, the text can be updated.
[18:28:59] <damien.saucez> luigi: we state already if it is easy do it
[18:29:19] <damien.saucez> luigi: we clearly state that the aim is not to protect the access to the router
[18:29:50] <damien.saucez> Chair: we have to be aware of the problem, but we probably not need to solve all them
[18:30:01] <damien.saucez> Chair: ok for me, I will ask to the mailing list
[18:30:24] <damien.saucez> Vince on stage
[18:31:16] <damien.saucez> lisp deployment
[18:31:23] <damien.saucez> wrong slide
[18:31:42] <damien.saucez> LISP-DDT
[18:31:58] <damien.saucez> DDT deployment and implementation
[18:32:02] <damien.saucez> not a tutorial
[18:32:05] <damien.saucez> slide 2
[18:32:48] <damien.saucez> will explain what has been discovered during experiments
[18:32:50] <damien.saucez> slide 3
[18:33:00] <damien.saucez> DDT = Delegated Database Tree
[18:33:14] <damien.saucez> terminology inspired by the DNS
[18:33:17] Eliot Lear leaves the room
[18:33:34] <damien.saucez> ALT was supposed to be easy to deploy, but in practice it was a pain in ***
[18:33:41] <damien.saucez> so moved to DDT
[18:34:14] <damien.saucez> tried with DNS but the implementation showed some difficulties
[18:34:18] <damien.saucez> slide 4
[18:34:22] <damien.saucez> two cisco implementation
[18:34:26] <damien.saucez> IOS and NXOS
[18:34:40] <damien.saucez> no progress on OpenLISP
[18:34:48] <damien.saucez> Verisign working on their own implem
[18:35:02] <damien.saucez> it is now deployed and has operational traffic on it
[18:35:05] <damien.saucez> it works
[18:35:20] <damien.saucez> does not include DDT-sec yet
[18:35:21] <damien.saucez> slide 5
[18:35:33] <damien.saucez> example of real configuration
[18:35:44] <damien.saucez> very similar to DNS (zones)
[18:36:40] <damien.saucez> the far right is always EID prefix
[18:37:21] <damien.saucez> slide 6
[18:38:33] <damien.saucez> a ddt server can't know if the prefix does not exist, instead it says: I don't know the prefix, and the resolver will try another server
[18:38:46] <damien.saucez> slide 7
[18:39:22] <damien.saucez> map server config is large (needs to know all its children), map resolver is simple (needs to know the roots)
[18:39:53] <damien.saucez> resolver and server can be combined on the same box
[18:39:57] <damien.saucez> slide 8
[18:40:05] <damien.saucez> scheme of the current deployment
[18:40:33] <damien.saucez> servers run different OSes and vendors
[18:41:07] <damien.saucez> hierarchy just as a proof of concept, not really needed as DDT is low overhead
[18:41:08] <damien.saucez> slide 9
[18:42:08] <damien.saucez> transition from ALT to DDT was in spring
[18:42:28] <damien.saucez> are looking right now for large deployment (structure?)
[18:42:31] <damien.saucez> slide 10
[18:42:37] <damien.saucez> open questions
[18:42:59] <damien.saucez> what "mapping provider eco-system"?
[18:43:12] <damien.saucez> we think the hierarchy will scale well
[18:43:17] <damien.saucez> (easy to replicate)
[18:43:46] <damien.saucez> would the "resource record" in the main DDT draft or a companion?
[18:44:34] <damien.saucez> open question: do we need separation?
[18:45:04] <damien.saucez> darrel to the mic
[18:45:45] <damien.saucez> <> is an open project, so anybody can register and propose to host DDT server/resolvers
[18:45:50] <damien.saucez> slide 11
[18:46:12] <damien.saucez> everything on the http::// <>
[18:46:27] <damien.saucez> ? on the mic:
[18:47:04] <damien.saucez> lookup of identities flat?
[18:47:43] <damien.saucez> vince: we looked at many mapping systems a time ago and we finally decided to follow a hierarchical approach as opposed to a flat
[18:48:10] <damien.saucez> vince: but as you have MR/MS you can plug whatever database infrastructure you want, even flat if you want
[18:48:19] <damien.saucez> dino:
[18:48:44] <damien.saucez> SDN is for network provisioning but we can see SDN to configure FIB
[18:49:30] <damien.saucez> dino: an app could program the mapping database directly
[18:50:35] <damien.saucez> dino: we thinik there is a scaling property even when you are roaming, you can always register to a MS
[18:51:42] <damien.saucez> ?? on mic: ok so what you need is only mapping flat
[18:52:12] <damien.saucez> dino: you could imagine a big MR/MS box and you program this box
[18:52:50] <damien.saucez> ??: ok, the mapping system is the controller then
[18:53:33] <damien.saucez> dino: logically we can see the mapping system as a centralized and operation wise it is distributed
[18:54:12] <damien.saucez> luigi: not much update but there is a MR/MS at paris6
[18:54:38] <damien.saucez> Vina: if scaling is not a matter, one server is ok
[18:55:05] <damien.saucez> Albert on stage
[18:55:22] <damien.saucez> lisp deployment
[18:55:39] <damien.saucez> slide 2
[18:56:22] <damien.saucez> give info about network elements introduced by LISP (xTR, MR, MS, pxTR...)
[18:56:37] <damien.saucez> technical content has not been updated much, mostly text and structure
[18:56:39] <damien.saucez> slide 3
[18:57:49] JoelHalpern joins the room
[18:58:07] <damien.saucez> slide 4
[18:58:17] <damien.saucez> map server and map resolvers
[18:58:59] <damien.saucez> slide 5
[18:59:07] <damien.saucez> proxy tunnel routers
[19:00:04] <damien.saucez> slide 6
[19:00:15] <damien.saucez> migration approaches to lis[
[19:00:26] <damien.saucez> s/lis[/lisp/
[19:01:57] <damien.saucez> a solution is to have an overlay of pxtr, advertised via BGP and the closest is taken
[19:02:07] JoelHalpern leaves the room: offline
[19:02:26] <damien.saucez> slide 7
[19:02:52] <damien.saucez> what a stub netadmin must do
[19:03:07] <damien.saucez> slide 8
[19:03:14] <damien.saucez> Acks
[19:03:17] <damien.saucez> last call?
[19:03:27] <damien.saucez> darel to the mic
[19:04:03] <damien.saucez> we have lisp proxy provider that have been deploying lisp, they moved customers from BGP to lisp
[19:04:14] <damien.saucez> they observed a reduction
[19:04:24] <damien.saucez> Joel: that is something useful to add to the document!
[19:04:58] <damien.saucez> joel: there is a real deployment and the result is significant and would improve the impact of the document
[19:05:23] <damien.saucez> Joel: DNS name or explicit RLOC
[19:05:29] <damien.saucez> darrel: slide 6
[19:05:36] <damien.saucez> slide 14
[19:06:02] <Florin Coras> but DNS names resolve to RLOCs or EIDs?
[19:06:56] <damien.saucez> this case makes sense if you have many pxtr; the second use, if you have settlement free peering but has to re-read the draft and emphases on this
[19:07:34] <damien.saucez> to rloc of map resolver
[19:08:05] <damien.saucez> chairs: now open mic
[19:08:42] <damien.saucez> we need to get this done! two docs intro and architecture
[19:08:55] <damien.saucez> does anyone has any feeling about structure of -intro-
[19:09:06] <damien.saucez> does it meet your expectations?
[19:09:08] <damien.saucez> darrel:
[19:09:23] Eliot Lear joins the room
[19:09:32] <damien.saucez> it meets for an implementer but not really for users
[19:10:32] <damien.saucez> joel: to some level it is ok, but if some section could be added for user, that would be indeed good, but we don't look to cover every deployment in -intro-
[19:10:36] <damien.saucez> albert:
[19:10:58] <damien.saucez> we must clarify what are the objective of this document
[19:11:31] <damien.saucez> chairs: people that will implement, so the doc to help you understand all the piece together
[19:11:50] <damien.saucez> albert: so this would be the first doc to read
[19:12:07] <damien.saucez> joel: 1. intro, 2. archi, 3 spec
[19:12:18] <damien.saucez> albert: so why not start with life of a packet
[19:12:37] <damien.saucez> chair: it has to be later in the document
[19:12:45] <damien.saucez> it will be in 6.1
[19:12:55] <damien.saucez> it is not written yet though
[19:13:07] <damien.saucez> albert: maybe early? 6 is too late
[19:13:23] <damien.saucez> chairs: yes, but hard to be before as there is first the terms
[19:13:39] <damien.saucez> albert: explain what then how or how then what?
[19:13:53] <damien.saucez> luigi: agree with albert
[19:14:08] <damien.saucez> luigi: lack of the big picture
[19:14:42] <damien.saucez> luigi: structure ok but writing gives me the filling that you have to already know the terminology to be able to read it
[19:14:57] <damien.saucez> luigi: we should ask people that don't know anything about lisp to see if they understand
[19:15:21] <damien.saucez> chairs: we would be happy to rearrange
[19:16:01] <damien.saucez> brian: we don't really know about lisp, but we know well the internet architecture so we will be able to give such feedback
[19:16:37] <damien.saucez> chairs: are there any section that someone does not agree it must be in the document
[19:17:28] <damien.saucez> darrel: more about addition than deletion. we might first look at the outline maybe
[19:17:47] <damien.saucez> chair: we are waiting for that!
[19:18:27] <damien.saucez> still many sections to be written
[19:18:37] <damien.saucez> 9 sections that need text
[19:18:48] <damien.saucez> are people in this room for those sections
[19:19:04] <damien.saucez> for the record: Darrel, Luigi, Albert, Vince, Damien
[19:19:49] <damien.saucez> luigi: add section about mapping system in general: the lisp architecture is open for new solutions
[19:19:56] <damien.saucez> chairs: YES
[19:20:28] <damien.saucez> sharon: local use would be nice to be increased
[19:20:57] <damien.saucez> dino: a lot of use cases not in the charter :-(
[19:21:25] <damien.saucez> chair: we should focus on what is in the charter, so just touch on them
[19:22:06] <damien.saucez> luigi: document with an overview of all the use cases, we started with scalability and figured out it would be use for other
[19:22:30] <damien.saucez> chairs: first finish the in charter use case!
[19:22:42] <damien.saucez> dino: but individual submission are allowed :-D
[19:23:16] <damien.saucez> dino: what precedence between this draft and docs in the queue?
[19:24:10] <damien.saucez> chair: no problem saying yes, it is not in the RFC but we know now that x would be better
[19:24:44] <damien.saucez> chair: what you think about arch?
[19:24:50] <Eliot Lear> hang on a sec
[19:25:12] <damien.saucez> sections to write in arch
[19:25:25] <damien.saucez> vince, darel, dino, vina, albert, luigi
[19:25:37] <Eliot Lear> mic: as i wrote on the mailing list, there needs to be slight clarification in the delineation between the core protocol, packet processing, and mapping system
[19:25:38] <damien.saucez> (many cookies )
[19:25:59] <damien.saucez> sharon: section on caching is too naive
[19:27:05] <damien.saucez> chairs: feel free to send text
[19:27:19] <damien.saucez> any more comment?
[19:27:30] <damien.saucez> dino: a lot of activity lvo3
[19:28:02] <damien.saucez> dino: we have to ongoing documents there
[19:28:09] <damien.saucez> dino: should be done here or there?
[19:28:12] <damien.saucez> chair: here
[19:28:31] <damien.saucez> chairs: but first make the current documents done!
[19:28:38] <Eliot Lear> by any chance would joel like the docs done?
[19:29:31] <damien.saucez> we don't really like the idea of moving protocols among WG.
[19:29:43] <damien.saucez> dino: use case vs spec
[19:30:03] <damien.saucez> chairs: use case there, spec changes here!
[19:30:06] <Eliot Lear> bye everyone
[19:30:11] <damien.saucez> done
[19:30:11] Eliot Lear leaves the room
[19:30:15] <damien.saucez> end of WG
[19:30:22] <damien.saucez> thank you
[19:30:25] <Florin Coras> thanks
[19:30:36] damien.saucez leaves the room
[19:30:41] Florin Coras leaves the room
[19:31:59] Luigi Iannone leaves the room
[19:33:06] damien.saucez joins the room
[19:33:38] Wassim M. Haddad leaves the room
[19:43:41] damien.saucez leaves the room
[19:46:13] damien.saucez joins the room
[20:03:27] damien.saucez leaves the room
[20:10:10] damien.saucez joins the room
[20:11:26] damien.saucez leaves the room
[20:26:13] Luigi Iannone joins the room
[20:28:41] damien.saucez joins the room
[20:28:57] damien.saucez leaves the room
[20:36:37] Luigi Iannone leaves the room
[20:36:47] Luigi Iannone joins the room
[20:36:56] damien.saucez joins the room
[20:40:37] damien.saucez leaves the room
[21:14:10] Luigi Iannone leaves the room
[21:15:55] Luigi Iannone joins the room
[21:47:23] Luigi Iannone leaves the room
[22:01:41] Wassim M. Haddad joins the room
[22:03:25] Luigi Iannone joins the room
[22:38:19] Wassim M. Haddad leaves the room
[22:46:04] Wassim M. Haddad joins the room
[22:51:43] Wassim M. Haddad leaves the room
[23:23:58] Luigi Iannone leaves the room
Powered by ejabberd Powered by Erlang Valid XHTML 1.0 Transitional Valid CSS!