IETF
sidr
sidr@jabber.ietf.org
Monday, March 11, 2013< ^ >
sandy has set the subject to: http://trac.tools.ietf.org/wg/sidr/trac/wiki/InterimMeeting20120929
Room Configuration
Room Occupants

GMT+0
[12:46:53] Wes George joins the room
[12:49:44] kasamatsu.kouhei joins the room
[12:56:22] danyork joins the room
[12:58:23] dseomn joins the room
[13:00:14] Sean Turner joins the room
[13:01:56] Stewart Bryant joins the room
[13:02:36] russmundy@jabber.org joins the room
[13:03:06] <Sean Turner> if anybody is listening to the audio stream let us know how the audio is when we get started
[13:03:54] <russmundy@jabber.org> I plan on being on the audio but so far nothing there
[13:04:43] <russmundy@jabber.org> link from the tools agenda web page == Not Found
The requested URL /ietf/ietf866.m3u was not found on this server.
[13:05:31] Juan-Pedro Cerezo Martin joins the room
[13:05:52] <Sean Turner> still nothing?
[13:06:10] <Wes George> I'll be scribing
[13:06:17] <russmundy@jabber.org> nope
[13:06:28] John Scudder joins the room
[13:07:03] weiler joins the room
[13:07:09] lepinski joins the room
[13:07:10] Arturo Servin  Ü joins the room
[13:07:45] <Sean Turner> working on it!
[13:07:55] <Sean Turner> the audio that is
[13:07:58] <Wes George> I'll try to scribe discussion until audio comes back
[13:08:02] bradd joins the room
[13:08:03] <russmundy@jabber.org> tks
[13:08:08] alexey.melnikov joins the room
[13:08:09] <Wes George> currently administrivia and note well
[13:08:10] <russmundy@jabber.org> thanks wes
[13:08:16] <Wes George> IPR discussion
[13:08:42] <Wes George> agenda bashing
[13:08:45] <Wes George> http://www.ietf.org/proceedings/86/slides/slides-86-sidr-6.pdf slide 7
[13:09:00] <Wes George> last minute change in what gets presented today vs tues
[13:09:18] <Wes George> both RPKI perf presentations today
[13:09:22] satoru.kanno@jabber.org joins the room
[13:09:41] <Wes George> tues - ruediger's presentation and relying party interop report from andrew chi
[13:09:48] markk joins the room
[13:09:57] <Wes George> any IEPG presentations reprised on Tues
[13:10:30] <Wes George> Matt Lepinski now presenting (sans slides) on BGPSec protocol spec draft
[13:10:51] <Wes George> currently this doc is waiting on requirements and threats to gain consensus
[13:10:57] <Wes George> to best off my knowledge, no open issues.
[13:11:05] <Wes George> may be add'l issues once we get reqs and threats nailed down
[13:11:18] <Wes George> new version vefore meeting to rewrite bgpsec capability neg
[13:11:45] <Wes George> prev version had both capability to advertise bgp speaker capable of sendingg bgpsec signed update msgs and a separate capaibilty indicating ability to receiev
[13:12:05] <Wes George> got feedback last mtg that this should be one capability - single capapbility with a bit in the firstr byte indicating send vs receive
[13:12:30] <Wes George> send an instance of this capability if you want to send, a different instance of it when you want to receive
[13:12:39] <Wes George> bit0 =send bit 1 = recv
[13:12:51] weiler leaves the room
[13:12:51] weiler joins the room
[13:12:51] <Wes George> capability into tthe session twice if you can send and receive
[13:12:55] <Wes George> singgle capabiliy with modal send-receive
[13:12:59] Jeffrey Haas joins the room
[13:13:07] weiler leaves the room
[13:13:19] <Wes George> making change required significant changes in supporting text in section 2
[13:13:20] <Wes George> but nott actually a significant change
[13:13:38] <Wes George> that's the only change in the doc, know of no other issues
[13:13:42] <Wes George> requesting reviews still - signifcant change since amsterdam interim
[13:13:56] <Wes George> questions?
[13:14:05] <Wes George> please put mic: in front of any comments you wish to be proxied
[13:14:20] wilton@jabber.isoc.org joins the room
[13:14:23] <Wes George> Sean turner discussing http://www.ietf.org/proceedings/86/slides/slides-86-sidr-8.pdf
[13:14:23] <Wes George> slide 1
[13:14:43] <Wes George> rtr keyingg - who has read
[13:15:25] <Wes George> comments - need to explain rationale for two approaches router vs operator
[13:15:43] <Wes George> I was worried about pki model that was built by a person, but here we're talking about routers and not people so that's ok, but need to clarify
[13:16:09] <russmundy@jabber.org> audio stream has appeared but is _very_ low in volume
[13:16:24] <Wes George> other comment was don't need to narrate all of router provisioning, thought I could point at another BcP, but that doesn't exist
[13:16:43] <Wes George> betteR?
[13:16:50] kivinen joins the room
[13:17:08] <russmundy@jabber.org> better but still low
[13:17:39] <Jeffrey Haas> In room is booming now, but still low on remote?  if so, we need an audio engineer to help.
[13:17:48] <Wes George> new version will incorporate currennt comments, but probably not actually done yet
[13:18:10] <Wes George> next slide bpsec algs
[13:18:18] <John Scudder> right. can someone on the audio stream confirm the audio is still unacceptable? if so I'll open a ticket.
[13:18:26] <russmundy@jabber.org> yes, remote is still very low - can only sort of hear voices
[13:18:31] <Wes George> IPR disclosure about ECDSA use in BGPSEc
[13:18:43] <russmundy@jabber.org> yes please open ticket
[13:18:57] <Wes George> sandy: seen other wgs presented with same IPR
[13:19:01] <Wes George> issues and discussions
[13:19:17] danyork leaves the room
[13:19:31] <Wes George> RFC6090 - used old references and showed how you could do eliptic curve with old references
[13:19:56] <Wes George> russ housley: as ietf chair : 6090 is a way to do these referencing only things that have been in the literature for 20+ years
[13:19:58] <Wes George> draw your own conclusions
[13:20:01] Dan York joins the room
[13:20:08] danyork joins the room
[13:20:13] <Wes George> there should be nothingg in the RFC about the patents- that all ggoes in the literature
[13:20:13] <Wes George> there's nothing you can say about it that will remain truie for the life of the RFC
[13:20:21] Wes George leaves the room
[13:20:21] <John Scudder> ticket opened
[13:20:36] <russmundy@jabber.org> tks
[13:20:48] hardaker joins the room
[13:21:44] <danyork> Wes just got booted out of the room, I'll be scribing until he returns
[13:22:21] <danyork> Done with Sean's presentation - no further questions
[13:22:39] <danyork> Next up - Carlos Martinez on "Multiple Publication Points"
[13:22:59] weiler joins the room
[13:23:17] <danyork> Draft: http://tools.ietf.org/html/draft-rogaglia-sidr-multiple-publication-points-02
[13:23:31] <russmundy@jabber.org> closer mike helps a lot
[13:23:46] <danyork> Slides: http://www.ietf.org/proceedings/86/slides/slides-86-sidr-3
[13:23:47] Wes George joins the room
[13:23:54] <danyork> Slide 2: The idea
[13:24:14] <danyork> Slide 3: Repository HA Options
[13:24:30] <Wes George> rewrote mostly via feedback from atlanta
[13:24:44] <Wes George> our current options are not satisfactory
[13:25:02] <Wes George> could wait for HTTP to mature, but not available now
[13:25:29] <Wes George> CDN doesn't provide you with DNS diversity -- still tied to single domain name
[13:25:43] <danyork> Slide 4: Progress since Atlanta
[13:25:52] <Wes George> DNS server farm - need a high-availability check, easier said than done
[13:25:52] <Wes George> slide 4
[13:26:18] markk leaves the room
[13:26:26] <Wes George> slide
[13:26:39] <Wes George> skipped slide 5, now on 6
[13:26:55] <Wes George> one could be HTTP, they are going to different domains
[13:27:17] <Wes George> you could even use literal IP address to be independent from DNS
[13:27:34] <Wes George> request WG adoptions
[13:27:42] <Wes George> end
[13:28:09] <Wes George> someone from JPNIC (diidn't get name)
[13:28:23] <Wes George> tried RPKI trial in Japan with operators and ISP
[13:28:33] <Wes George> availability of repository is a big issue
[13:28:38] <danyork> Taiji Kimura from JPNIC
[13:28:43] <Wes George> I think your document will cover some of those single points off failuire
[13:29:02] <Wes George> we have question about difference between multiple repository
[13:29:09] <Wes George> inconsistencies between them?
[13:29:22] <Wes George> a: doc doesn't address, but needs to be analysed
[13:29:31] <Wes George> every repository consists of 2-3 copies, problem will happen today
[13:29:35] <Wes George> not specific to this solution
[13:29:53] <Wes George> rob austein: not a friend off this draft - any replication problem will be there regardless
[13:30:16] <Wes George> now discussing http://www.ietf.org/proceedings/86/slides/slides-86-sidr-0.pdf
[13:30:19] <Wes George> PQ of the CP
[13:30:23] <Wes George> andy newton presenting
[13:30:36] <Wes George> putting CPS into res crts
[13:30:40] <Wes George> slide 2
[13:30:47] <Wes George> slide 3
[13:30:55] <danyork> Draft: http://tools.ietf.org/html/draft-newton-sidr-policy-qualifiers-01
[13:31:06] <Wes George> slide 3
[13:31:23] <Wes George> current RFC ambiguouos as to if this is allowed
[13:31:43] <Wes George> 1 of the 3 validators accepted CPS url, the other 2 didn't
[13:31:48] <Wes George> one line change to fix, now all do
[13:31:50] <Wes George> slide 4
[13:32:22] <Wes George> slide 5
[13:32:25] <Wes George> since- 01
[13:33:27] <Wes George> rob austein: other than not regecting the data, there's nothing that you want the RPs to do with this?
[13:33:31] <Wes George> correct
[13:33:52] <Wes George> ruediger - we do this, make the lawyers happy, and ARIN removes the RPA?
[13:34:12] <russmundy@jabber.org> now, as long as speakers are close to the mike, remote audio is now satisfactory
[13:34:20] <Wes George> john curran: possible an org might want to look at the URLs for the certs it relies on , but that's up to them
[13:34:46] <Wes George> this provides an addititonal reassurance in web of protection
[13:34:59] <Wes George> if community were to use methods of distribution with multipole things off the TAL, this might help
[13:35:32] <Wes George> ruediger - what are we actually going to do for having a valid communication of requirements to useres
[13:35:44] <Wes George> we have to publish the stuff htat is relevant for the general reliing party here
[13:35:55] <Wes George> we have to put the binding rules into some public general doc
[13:36:10] <Wes George> not rely on having every CA develop its own specific stuff and asking the world to take care of it
[13:36:24] <Wes George> or if we put this in somewhere, and no one cares, I'm happy
[13:36:47] <Wes George> curran: if you're talking about general liabilitty, bringin those to IETF isn't relaly constructive
[13:37:06] <Wes George> track record dealing with thoes in technincal forum instead of user community is pretty low
[13:37:18] <Wes George> george michealson : not bash on arin - APNIC needs this too
[13:37:26] <Wes George> we expect to include this class of pointe rin our TAL
[13:37:26] <John Scudder> Re audio, Alexa sez "John,
I ran into someone from the NOC in the hallway and he thought this had been fixed. Can
you confirm this true?
Thanks,
Alexa"
[13:37:38] <John Scudder> Can someone comment?
[13:37:51] <Wes George> ross said it was better
[13:38:06] <Wes George> russ
[13:38:29] <Wes George> morrow: two issues being discussed, clarify
[13:38:39] <Wes George> george - not RPA, discussing the draft
[13:38:48] <Wes George> sean turner - I think this is less horrible now
[13:38:55] <Wes George> terry manderson - no harm, let's do it
[13:39:00] <Wes George> asking for WG adoption
[13:39:04] lepinski leaves the room
[13:39:11] <russmundy@jabber.org> a little more volume might be nice but not essential
[13:39:32] <Wes George> now discussing http://www.ietf.org/proceedings/86/slides/slides-86-sidr-7.pdf
[13:39:41] <Wes George> George Michaelson presenting
[13:39:51] <Wes George> cut down preso from APRICOT
[13:39:55] <Wes George> slide 2
[13:40:03] lepinski joins the room
[13:40:21] <Wes George> we did some changes that we didn't communicate well
[13:40:21] <Wes George> slide 4 - RPKI TA Framework
[13:40:43] <Wes George> slide 5
[13:42:06] <Wes George> slide 6 - anchor
[13:42:12] <Wes George> slide 7 - apnic's TA changes
[13:42:26] <Wes George> slide 8 - ta changes
[13:42:40] <Wes George> slide 9 - ta changes
[13:43:13] <Wes George> slide 10
[13:43:23] <Wes George> "re-align..."
[13:43:50] <Wes George> slide 11 - ta changes diagram
[13:44:15] <Wes George> slide 12
[13:44:15] <Wes George> slide 13
[13:44:24] <Wes George> "this inter-rir registry..."
[13:44:52] <Wes George> slide 14 "by converting..."
[13:45:50] <Wes George> slide 15 - "other rirs..."
[13:47:18] <Wes George> slide 15 "we would like to understand..."
[13:47:28] <Wes George> slide 17
[13:47:39] <Wes George> slide 18 provisioning protocol services
[13:48:18] <Wes George> slide 20 "we have written a simple..."
[13:48:41] <Wes George> slide 21 "by incorporatingg..."
[13:48:54] markk joins the room
[13:48:57] <Wes George> slide 22 "we believe..."
[13:50:30] <Wes George> rob austein: george and I discussed taking this to PKIX instead of here
[13:51:08] <Wes George> george is correct that I'm working on a draft based on feedback on trying to deploy the original version... people said "what a piece of... fix this!"
[13:51:21] <Wes George> think I finally have a way forward with this, will be submitting a draft RSN
[13:51:30] <Wes George> andy newton - you should bring draft to this WG. Should make it a standard
[13:51:46] <Wes George> sean turner: PKIX won't accept, bring it here
[13:52:07] <Wes George> sandy: are we talking about an existing drafT?
[13:52:11] <Wes George> rob : no this is a -00
[13:52:51] <Wes George> taji from JPNIC - may need same things in LIR, may need own sub-ca
[13:53:05] <Wes George> george: transfers have to pass through their parent
[13:53:25] <Wes George> in that sense you will have to dialog via APNIC for transfers for the signing material to be valid
[13:53:30] <Wes George> we'd have to issue certs to make that happen
[13:53:47] <Wes George> you're free to construct local trust anchor/TAL and include that in your TALSet as relying party
[13:54:03] sandy joins the room
[13:54:06] <Wes George> but in general sense, the community benefit is to have this go through the parent
[13:55:33] <Wes George> rob austein: he was asking wheteher or not they need separate TAs.
[13:55:45] <Wes George> george: all direct allocations rest within APNIC
[13:56:24] <Wes George> steve kent: this same reasoningg regarding JPNIC and transfers - move it up a level, having multiple TALs for APNIC to reflect the providence of allocations is probably not a good idea
[13:56:50] <Wes George> if we're looking to gget to IANA as global TA, the transfers between RIRs would be handled up, and down just as you described, and the sepearate TAs wouldn't be there
[13:57:15] <Wes George> IANA would issue resources into that to reflect the transfer from another RIR
[13:58:51] <russmundy@jabber.org> wow - the remote audio just went up, yea, it's now easy to hear - thanks
[13:59:04] <danyork> Arturo Servin from LACNIC at the mic
[13:59:18] <danyork> John Curran at the mic
[13:59:18] Andrew Chi joins the room
[13:59:37] <Wes George> john curran: we're doingg a lot of work to figure out how to make transfers work at the /8 level, but nothingg smaller
[13:59:49] <Wes George> unless you're doing that under APNIC, there's more thought needed
[14:00:01] <Jeffrey Haas> I'm failing to follow enough of the conversation to take useful notes.  Help appreciated.
[14:00:07] <Wes George> how do we do this between an RIR like ARIN to a national RIR like JPNIC
[14:00:23] <danyork> Andrei Robachevsky from Internet Society at the mic
[14:00:41] <Wes George> andrei: this TA will be signed by ARIN, preparation for proper heierarchy
[14:01:13] <Wes George> george: three modes operation - assert 0/0
[14:01:27] <Wes George> assert the list of things you know you have - everything you have is valid, no claims about what you don't have
[14:01:42] <Wes George> third - split TAL, vesting chain
[14:02:07] <Wes George> can't have a single cert that covers everythingg that is co-signed by other RIRs
[14:02:22] <Wes George> this exposes 5 discrete TALs
[14:02:35] <Wes George> we named them APNIC from ARIN, APNIC from RIPE, APNIC from AFRINIC
[14:02:40] <Wes George> that was our sense of where we were going
[14:02:52] <danyork> Rob Austein at the mic
[14:03:22] <Wes George> rob: my understanding of 3779 is agnostic of these relationships, you can represent any of this
[14:03:34] <Wes George> assumption was that one would want th e smallest resource tree
[14:03:42] <Wes George> leastt number of certs, direct chain, smallest tree
[14:04:04] <Wes George> for JPNIC, might actually swing the /16 at the IANA level
[14:04:17] <Wes George> sounds like RIRs only deal with /8s, so "get over it"
[14:04:28] <Wes George> they have tooling that all focuses on /8
[14:05:11] <Wes George> george: in operating cert sense, there may be 6 (includingg RFC1918 and multicast space, which RIRs don't have purview over)
[14:05:59] <Wes George> ruediger volk: Is the question of how transfers are handled just a private matter between RIRs and IANA
[14:06:12] <Wes George> or should it be publicly documented/discussed as it is reflected in RPKI
[14:06:18] <Wes George> george: this isn't protocol or technology
[14:06:35] <Wes George> not saying "should/should not be discussed" but can't see it being on the plank for SIDR
[14:06:47] <Wes George> ruediger: i appreciate the idea of exposing what is done operationally
[14:07:07] <Wes George> participants in the system need to understand the context they're operating in
[14:07:21] <Wes George> valid forum here for makingg the whole system resilient and reliable
[14:07:51] <Wes George> george: registry view, natural comparison : lands registry - how do they work?
[14:10:35] <Wes George> how do they migrate, where is authority
[14:10:36] <Wes George> properties of the address field chopped up like land subdivision
[14:10:50] <Wes George> when we put 5 TALs out there, that was a strange moment for some people
[14:10:50] <Wes George> getting a vibe that we maybe were out in the field compared to otheres
[14:11:14] Wes George leaves the room
[14:11:33] Karen O'Donoghue joins the room
[14:12:30] Wes George joins the room
[14:13:12] <Wes George> sorry, missed a bit due to jabber hiccup
[14:13:29] <Wes George> sorry, missed a bit due to jabber hiccup
[14:13:29] <Wes George> Mark Kosters is working on a draft that will deal with transfers
[14:14:21] <Wes George> sandy - issue with operators deciding who to trust and who not to , hard to keep up with, referenced KC Claffey message on list
[14:14:38] <Wes George> george: trying to explain what we believe we know about it
[14:14:52] <Wes George> this may be a NRO question to be honest
[14:15:25] <Wes George> addresses have become nat'l strategic matters - are regulated
[14:15:30] <Wes George> political reality drives some changes from just technical standards
[14:15:30] <danyork> George: these are higher level policy issues that are handled by governments in some cases
[14:15:42] <Wes George> vietnam requires permission ffrom gov't to transfer out of their region
[14:15:59] <danyork> To use IETF-speak, these are Layer 8/9 issues :-)
[14:16:33] <Wes George> rob austein
[14:16:47] <danyork> Anyone else finding a bit of lag to posts to the jabber server?
[14:16:47] <Wes George> response to sandy's q
[14:16:47] <Wes George> the TAL doesnn't include resources
[14:17:07] <Wes George> point of the TAL is not to include - used self-signed cert
[14:17:20] <Wes George> one can retrieve the TAs and look for overlaps, who is claiming what
[14:17:20] <Wes George> two priblems - one claims 0/0 - overlap
[14:17:41] <Wes George> larger problem - what is the RP supposed to do about this? either you trust the TA or you don't
[14:17:57] <Wes George> yes, more than a minute
[14:18:27] <Wes George> back to slides, no idea which number - provisioningg protocl services "APNIC has also identified..."
[14:19:14] <Wes George> comments?
[14:19:21] <Wes George> anything in the jabber room?
[14:19:29] <danyork> Sean Turner at mic
[14:19:29] <Wes George> sean turner: are you basically talking about putting things on hold
[14:19:30] <Wes George> george: nont in the sense that would be crypto visible
[14:19:46] <Wes George> just UI
[14:19:53] <Wes George> if you say you're moving and we do it too soon, everything disappears
[14:20:27] <Wes George> make it explicit "I'm holding this..." then when you're happy, you kill the old one
[14:20:27] <Wes George> chris morrow: sounds like changing stuff in DNS - you don't pull old servers til the new ones are there
[14:21:55] <Wes George> now discussing http://www.ietf.org/proceedings/86/slides/slides-86-sidr-2.pdf
[14:21:56] <Wes George> Oleg Muravskiy
[14:21:56] lepinski leaves the room
[14:22:04] <Wes George> slide 2
[14:22:22] lepinski joins the room
[14:22:58] <Wes George> slide 3
[14:23:07] Andrew Chi leaves the room
[14:24:24] weiler leaves the room
[14:24:31] lepinski leaves the room
[14:24:33] <Wes George> slide 5
[14:24:54] weiler joins the room
[14:25:42] <Wes George> slide 6
[14:25:48] lepinski joins the room
[14:26:36] <Wes George> slide 7
[14:26:38] Andrew Chi joins the room
[14:28:26] <Wes George> steve kent: this is true, but any of the other thigns proposed will not solve this
[14:28:36] <Wes George> person runnin the software can mess up and build a bad manifest
[14:28:50] <Wes George> rob austein: what steve said is correct -program error
[14:29:05] <Wes George> rsync has no libraries, you have to go through filesystem and deal with objects
[14:29:17] <Wes George> differes by distribution as to what the underlying filesystem is doing
[14:29:54] <Wes George> steve: what we're talking about is that only one off these two issues will be addressed by switching to someothing other than rsync
[14:30:07] John Scudder leaves the room
[14:30:16] <Wes George> the requirements doc is not well-written
[14:30:17] <Wes George> it says " I already know what the answer is, let's try to arrive there"
[14:30:46] weiler leaves the room
[14:30:50] weiler joins the room
[14:31:39] sandy joins the room
[14:31:59] <Wes George> oleg - proposal for new protocol to let client calculate what it neesd from server instead of server figuring out for client
[14:32:22] John Scudder joins the room
[14:32:38] <Wes George> steve: agree that we should explore other distributions, but not with bad draft to justify
[14:33:33] <Wes George> george: I think there is a problem, but more related to filesystem and what you publish publicly
[14:33:50] <Wes George> even if you do best hierarchicy, large period of incomplete state during periodic update
[14:34:04] <Wes George> likllihood of that being visible publicly is quite high
[14:34:05] <Wes George> approaches to minimize this,
[14:34:26] <Wes George> like the idea of defining somethingg better
[14:35:10] gerardo joins the room
[14:35:33] <Wes George> slide 8
[14:35:34] <Wes George> new req's
[14:36:20] gerardo leaves the room
[14:36:43] Juan-Pedro Cerezo Martin leaves the room
[14:37:21] weiler leaves the room
[14:39:01] gerardo joins the room
[14:39:05] <Wes George> slide 9
[14:39:51] <Wes George> steve kent: a lot of these are highly questionable assumptions
[14:40:17] Juan-Pedro Cerezo Martin joins the room
[14:40:17] <Wes George> eric osterweil: these numbers look close to the numbers we arrived at independently
[14:40:23] <Wes George> steve kent no I think that he took those numbers
[14:40:23] <Wes George> eric - no citation
[14:40:50] <Wes George> all of the above should be read in the passive aggressive font
[14:40:50] <Wes George> ;-)
[14:41:06] <Jeffrey Haas> s/passive//
[14:41:48] <John Scudder> ^
[14:42:19] <Wes George> slide 10
[14:42:24] Nico joins the room
[14:43:23] Nico leaves the room
[14:43:40] Nicolás Fiumarelli - LACNIC joins the room
[14:44:29] <Wes George> slide 1
[14:44:30] <Wes George> slide 11...
[14:45:31] <Wes George> rob austein: is it a requirement that you have these multiple files
[14:45:36] <Wes George> a: not necessarily
[14:45:50] <Wes George> slide 12
[14:46:00] sandy leaves the room
[14:46:12] <sandy> (hope the strong hum we are hearing here in the room is not coming thru the audio stream)
[14:46:33] <sandy> (please report if it is)
[14:47:01] <Jeffrey Haas> the cabling for the portable mic is running over the power cables.
[14:47:04] <John Scudder> I would like to report that the strong hum is audible in the room and is mighty annoying :-P
[14:47:06] <wilton@jabber.isoc.org> (whatever someone did a few moments ago to stop the hum… it would be great if they did it again! ;^(
[14:47:15] <Wes George> slide 13
[14:47:18] <Sean Turner> here we go ...
[14:48:07] <Wes George> slide 14
[14:48:13] <Sean Turner> mic switch
[14:48:34] weiler joins the room
[14:49:25] Wes George leaves the room
[14:50:09] <danyork> Wes had to drop off for a moment, I'll scribe
[14:50:23] Wes George joins the room
[14:50:43] <danyork> Sriram Kotikalapudi (from NIST) at mic.
[14:50:47] <Wes George> terry manderson: size component in the XML?
[14:51:08] <danyork> Going back to the slide about scalability of rsync server
[14:51:32] <danyork> George Michaelson at mic
[14:51:50] <danyork> Rob Austein was at mic to make same point as George
[14:51:54] <danyork> John Scudder at mic
[14:51:57] <Wes George> john scudder: there's no potential for intermediate state compression?
[14:52:12] <Wes George> if object changes from 1-3, you serve 2 deltas, correct?
[14:52:12] <Wes George> yes
[14:52:24] <Wes George> eric osterweil : good points, nuanced points
[14:52:31] <Wes George> would help if they were more highlighted
[14:52:35] <danyork> Eric (?) at mic
[14:52:45] <Wes George> issues you brought up pointing out RSYNC causing issues, people responded that this wasn't a rsync specific problem
[14:52:57] <Wes George> documenting engineering tradeoffs would be useful
[14:52:57] <danyork> Rob Austein at mic
[14:53:10] <Wes George> rob austein - semantics of protocol are familiar
[14:53:16] <Wes George> go back a couple of ancestors, dns incremental zone transfers
[14:53:23] <Wes George> they did version compression
[14:53:30] <Wes George> going from whatever version you have to current version
[14:53:49] <Wes George> understand why you specified, but would be good to leave open the option to do this
[14:53:57] <Wes George> oleg - is there any reason to keep the intermediate version
[14:54:18] <Wes George> rob: suggested that to tim, he was concerned about load on publication engine
[14:54:18] <Sean Turner> I'd like to echo what eric said at the mic about documenting the engineering tradeoffs
[14:54:52] <Sean Turner> I mean hasn't this kind of thing been done before
[14:55:19] <Wes George> now, sriram presenting http://www.ietf.org/proceedings/86/slides/slides-86-sidr-1.pdf
[14:55:44] <Jeffrey Haas> and the lesson is usually "don't use a file system, or at least a canned one".  See nntp servers of yore.
[14:56:09] <Wes George> slide 2
[14:56:29] <Wes George> aside: wait, when did we start putting exec summaries in IETF slides?
[14:56:40] sandy joins the room
[14:58:02] <danyork> Wes George: maybe when there's a deck with 29 slides?  :-)
[14:58:02] <Wes George> slide 3
[15:00:17] <Wes George> slide 4
[15:02:42] <Wes George> slide 5
[15:03:44] <Wes George> slide 6
[15:04:39] <Wes George> slide 7
[15:04:42] Sean Turner leaves the room
[15:05:21] weiler leaves the room
[15:05:22] weiler joins the room
[15:05:56] <Wes George> slide 9
[15:07:30] <Wes George> shane amante: question about numbers respect sync time per object
[15:07:36] <Wes George> doe sthis include propagation delay?
[15:07:37] <Wes George> sriram: yes does include
[15:07:49] <Wes George> chris - do you mean bgp propagation or disk to disk
[15:07:53] <Wes George> shane - the latter
[15:07:56] <Wes George> slide 10
[15:08:11] markk leaves the room
[15:08:37] weiler leaves the room
[15:08:38] weiler joins the room
[15:08:46] <Andrew Chi> to address shane's question: the BBN numbers came from total wall-clock time of an actual rsync between BBN and RIPE a few weeks ago
[15:08:49] <Wes George> slide 11
[15:09:22] <Wes George> slide 12
[15:09:30] Stewart Bryant leaves the room
[15:09:37] kasamatsu.kouhei leaves the room
[15:10:18] kasamatsu.kouhei joins the room
[15:10:24] <Wes George> george: you modeled this on 2 vms sharing the same hardware?
[15:10:30] <danyork> George Michaelson at the mic
[15:10:42] <Wes George> but this is probably not an apporpriate fabirc for testing the behavior
[15:10:45] <Wes George> slide 14
[15:10:46] Sean Turner joins the room
[15:10:51] <danyork> Testing jabber lag
[15:10:55] <Wes George> oops slide 13
[15:11:08] <John Scudder> George's comment should be in the dripping-with-irony font.
[15:11:15] <Wes George> dan I reported the jabber lag to MTD
[15:11:36] <Wes George> forget unicode, we need intent fonts
[15:11:45] <John Scudder> same-day-service for jabber messages
[15:12:04] <Wes George> yet another to add to my neverending list of april fools RFCs that I'll never write
[15:12:15] <danyork> Wes George: So did I... just looking to time it.  It's taking 8-10 seconds for my messages to appear back in my window.
[15:12:37] <Wes George> slide 154
[15:12:38] <Jeffrey Haas> http://www.collegehumor.com/article/6872071/8-new-and-necessary-punctuation-marks
[15:12:49] <Wes George> sigh... fat fingers. slide 15
[15:12:56] <Wes George> jeff, yea seen that
[15:14:10] <Wes George> slide 16
[15:15:21] <Wes George> slide 17
[15:16:51] <Wes George> slide 18
[15:18:08] markk joins the room
[15:18:25] russmundy@jabber.org leaves the room
[15:18:42] <Wes George> slide 19
[15:19:17] russmundy@jabber.org joins the room
[15:19:47] <Wes George> slide 20
[15:21:09] <danyork> George M. at mic again
[15:21:12] <Wes George> george: 19 and 20 are thehse assertions or was this experimentally measured?
[15:21:24] <Wes George> was it measuered, or modeled
[15:21:46] <danyork> Steve Kent at mic - it's a model
[15:21:46] <Wes George> sri: this is a model
[15:22:17] <Wes George> steve: it's a model, but uses a per-object fetch time based on longer times than reality, so it's conservative
[15:22:19] <Wes George> slide 21
[15:22:46] <Wes George> slide 22
[15:23:05] <Wes George> slide 23
[15:23:35] <Wes George> george: good piece of work. can critique, but won't do at mice
[15:23:35] <Wes George> mic
[15:23:52] <Wes George> couple things of concern - more instances of address holder not being the AS holder
[15:23:54] dseomn leaves the room
[15:24:10] <russmundy@jabber.org> pls ask george to speak up
[15:24:21] <russmundy@jabber.org> or get closer to mike
[15:24:24] <russmundy@jabber.org> mic
[15:24:30] <Wes George> roa count is not same as AS count, because AS holder and address holder count diverge
[15:25:18] <Wes George> suspect the total count is double your estimate, 3-4k in asia alone
[15:25:38] <Wes George> some people have 4000 direct children, no improvement for hierarchy
[15:25:56] <Wes George> personal exp operating Rsync with large communities of people
[15:26:06] SwedeMike joins the room
[15:26:09] <Wes George> optimistically, you just have to walk tree and return outcome
[15:26:21] Andrew Chi leaves the room
[15:26:25] <Wes George> 400 people/hr have to go through, 10 bindings/sec
[15:26:36] <Wes George> not like web where you do a binding and they're gone
[15:26:38] <Wes George> these persist
[15:26:51] <Wes George> you have to walk every file in your filsystem in rcync to discover if it's changed
[15:27:19] <Wes George> the host component (modeled rather than measured) during this sort of load, and it changes the case such that delay will be much higher as a host cost
[15:27:30] david joins the room
[15:27:43] <Wes George> bad assumption for this to be linearly scaled assuming well-behaved
[15:27:51] <Wes George> some parts look optimistic
[15:28:16] <Wes George> 30-50 rsyncs destroying your filecache because they're looking at different parts of teh file system
[15:28:29] <Wes George> steve: when we're doing this, we're doing it against an unloaded server
[15:28:37] <Wes George> george: optimistc sense of how good it could be
[15:28:45] <Wes George> steve: didn't understand your comment on pub points
[15:28:56] <Wes George> george: about your fanout
[15:29:17] <Wes George> there's 5, bbut underneath us, we have 5000. not a neat tree that descends in biaries
[15:29:20] weiler leaves the room
[15:29:20] weiler joins the room
[15:29:55] <Wes George> steve: took total number of ASNs as surrogate for resource holder - a large number will outsource. the others will be spread out between RIRs and between non-stub ASNs
[15:30:00] Andrew Chi joins the room
[15:30:03] <Wes George> we've accomodated the resource holders
[15:30:28] <Wes George> APNIC region has NIRs in any significant numbers, may be noise relative to the total we've spread out
[15:30:34] <Wes George> might not be as uniformly distributed
[15:30:53] kasamatsu.kouhei leaves the room
[15:31:18] kasamatsu.kouhei joins the room
[15:31:44] Karen O'Donoghue leaves the room
[15:31:56] <Wes George> sandy : INTERRUPT
[15:31:57] <Wes George> out of time
[15:32:01] markk leaves the room
[15:32:02] bradd leaves the room
[15:32:28] <Wes George> eric osterweil: applaud independent measurement, challenge othehr work
[15:32:40] weiler leaves the room
[15:32:41] weiler joins the room
[15:32:48] <Wes George> danger in being too presumptive about things you can't measure and then modeling on top of that
[15:33:06] <Wes George> once you get a bunch of assumptions you build your measurement on, it starts becoming hard to model accurately
[15:33:22] <Wes George> we have data from real fetches, posted on our site
[15:33:26] <Wes George> they're higher than yours
[15:34:21] Dan York leaves the room
[15:34:27] danyork leaves the room
[15:34:47] weiler leaves the room
[15:34:48] weiler joins the room
[15:34:50] Sean Turner leaves the room
[15:35:00] <Wes George> as of 5 min ago, RIPE's throughput 0.69 object/sec in single,
[15:35:48] <Wes George> steve kent: numbers are measurements from repositories, not synthetic data
[15:35:57] <Wes George> but different implementations
[15:36:03] <Wes George> eric: rsync is rsync
[15:36:11] <Wes George> steve: bad implementations of RP software....
[15:36:14] weiler leaves the room
[15:36:17] <russmundy@jabber.org> I guess what happens when you go over time - the audio gets dropped :-)
[15:36:48] <Wes George> steve: unlikely to do key per router (c.f. MD5)
[15:36:52] <russmundy@jabber.org> Thanks a LOT to Wes George for taking such good jabber scribe notes
[15:36:57] <Jeffrey Haas> ^
[15:37:05] <Wes George> danny : disagree
[15:37:21] <Wes George> morrow: two separate security zones, hostile vs doesn't
[15:37:30] <Wes George> may be that we have individual keys in the hostile, shared keys in non-hostile
[15:37:36] <Wes George> haven't had discussion internally
[15:37:53] <Wes George> but everyone may make a different decision here
[15:38:00] <Wes George> not sure what the right answer is as far as model is concerned
[15:38:09] <Wes George> stating what model is and why you got numbers is still good
[15:38:29] <Wes George> shane: as an operator of several thousand routers, there's no way I'm going to have a single router key for entire AS
[15:38:38] alexey.melnikov leaves the room
[15:38:46] <Wes George> steve: and this is the same as your MD5 values today?
[15:38:57] <Wes George> shane: no, these are managed on a per-session basis
[15:39:44] hardaker leaves the room
[15:39:49] sandy leaves the room
[15:42:24] lepinski leaves the room: Computer went to sleep
[15:42:46] Jeffrey Haas leaves the room
[15:42:49] Wes George leaves the room
[15:42:56] Arturo Servin  Ü leaves the room
[15:43:03] Andrew Chi leaves the room
[15:43:30] sandy leaves the room: Computer went to sleep
[15:43:36] david leaves the room
[15:44:50] hardaker joins the room
[15:45:08] hardaker leaves the room
[15:45:41] glen joins the room
[15:45:46] <glen> Test
[15:45:56] <glen> Sorry for the intrustion, testing this room, please ignore.
[15:46:14] satoru.kanno@jabber.org leaves the room
[15:46:23] glen leaves the room
[15:46:54] kasamatsu.kouhei leaves the room
[15:47:12] Sean Turner joins the room
[15:47:46] wilton@jabber.isoc.org leaves the room
[15:49:47] Sean Turner leaves the room
[15:49:56] Sean Turner joins the room
[15:51:06] russmundy@jabber.org leaves the room
[15:51:45] gerardo leaves the room
[15:51:51] Nicolás Fiumarelli - LACNIC leaves the room
[15:52:00] Juan-Pedro Cerezo Martin leaves the room
[15:53:17] John Scudder leaves the room
[16:00:34] kivinen leaves the room
[16:02:57] wilton@jabber.isoc.org joins the room
[16:05:39] Karen O'Donoghue joins the room
[16:12:55] wilton@jabber.isoc.org leaves the room: Replaced by new connection
[16:12:56] wilton@jabber.isoc.org joins the room
[16:15:43] Karen O'Donoghue leaves the room: Replaced by new connection
[16:15:43] Karen O'Donoghue joins the room
[16:16:08] Arturo Servin  Ü joins the room
[16:22:00] Arturo Servin  Ü leaves the room
[16:22:57] hardaker joins the room
[16:27:16] <SwedeMike> seems to work.
[16:33:35] Karen O'Donoghue leaves the room
[16:49:23] lepinski joins the room
[16:49:42] lepinski leaves the room
[16:50:36] wilton@jabber.isoc.org leaves the room
[16:53:17] danyork joins the room
[16:54:03] Dan York joins the room
[16:54:17] danyork leaves the room
[16:54:30] Dan York leaves the room
[16:59:22] Stewart Bryant joins the room
[17:02:09] Sean Turner leaves the room
[17:04:12] Karen O'Donoghue joins the room
[17:04:24] Karen O'Donoghue leaves the room
[17:06:27] hardaker leaves the room
[17:12:04] satoru.kanno@jabber.org joins the room
[17:16:59] satoru.kanno@jabber.org leaves the room
[17:25:55] John Scudder joins the room
[17:26:20] John Scudder leaves the room
[17:40:16] markk joins the room
[17:40:26] markk leaves the room
[17:55:05] hardaker joins the room
[18:04:31] wilton@jabber.isoc.org joins the room
[18:04:45] wilton@jabber.isoc.org leaves the room
[18:44:31] hardaker leaves the room
[18:44:31] hardaker joins the room
[19:31:46] hardaker leaves the room
[19:32:01] Stewart Bryant leaves the room
[19:48:10] hardaker joins the room
[20:12:51] sandy joins the room
[21:11:33] hardaker leaves the room
[21:35:03] sandy leaves the room
[21:50:58] hardaker joins the room
[21:51:20] sandy joins the room
[23:26:52] sandy leaves the room
[23:46:50] hardaker leaves the room
Powered by ejabberd Powered by Erlang Valid XHTML 1.0 Transitional Valid CSS!