[07:43:20] Alissa Cooper joins the room
[07:43:32] Alissa Cooper leaves the room
[07:44:54] Jack joins the room
[07:51:04] sebastian.kiesel joins the room
[07:51:28] Jan.Seedorf joins the room
[07:51:30] saverio.niccolini joins the room
[07:51:36] Jan.Seedorf leaves the room
[07:53:24] brabson joins the room
[07:53:26] lairdp joins the room
[07:53:37] Alissa Cooper joins the room
[07:53:58] <lairdp> hi, everyone.
[07:54:50] Jan.Seedorf joins the room
[07:54:59] <saverio.niccolini> hi
[07:55:16] <sebastian.kiesel> howdy
[07:55:23] <Jan.Seedorf> hi
[07:55:48] jiangxingfeng joins the room
[07:56:51] <saverio.niccolini> anybody able to connect to ietf72-ch6 streaming? the web site says: Convention 3
channel-6 recorder appears to have a bad disk,
 currently suffering an outage
[07:56:53] kazu joins the room
[07:57:20] psavola joins the room
[08:00:17] jlcjohn joins the room
[08:00:33] Damon joins the room
[08:00:49] <lairdp> here we go!
[08:01:31] <lairdp> /scribe me. :-)
[08:01:47] Damon leaves the room
[08:01:58] victor.pascual joins the room
[08:02:02] saverio.niccolini leaves the room: Replaced by new connection
[08:02:47] csp joins the room
[08:03:13] CMW5S41 joins the room
[08:03:17] Haurhiko Nishida joins the room
[08:03:22] <jlcjohn> I'm getting no audio on channel 6
[08:03:22] nm joins the room
[08:03:36] <lairdp> Agenda. There's a lot of work being done in this area, so we're taking a survey of the existing work.
[08:03:53] xiaohunhun joins the room
[08:04:02] <lairdp> Ending in charter discussion.
[08:04:08] <lairdp> No agenda additions from the floor.
[08:04:10] xavi.mila joins the room
[08:04:18] Damon.Wischik joins the room
[08:04:19] <lairdp> Problem stratementL
[08:04:43] saverio.niccolini joins the room
[08:04:44] nov joins the room
[08:05:08] <saverio.niccolini> it seems audio streaming not working
[08:05:14] <lairdp> Internet applications evolve over time: email to file transfer to web browsing to p2p.
[08:05:40] <lairdp> Issues in p2p traffic:
[08:05:51] jinhong joins the room
[08:06:12] <lairdp> - 50-85% of total traffic. upstream as well as downstream. bandwidth-greedy. interferes with real-time traffic. unpredictable.
[08:06:58] shikob joins the room
[08:07:05] <lairdp> p2p has been in the news e.g. Comcast/BitTorrent, BBC iPlayer, then collaboration (e.g. P4P)
[08:07:25] <lairdp> IETF P2P Infrastructure Workshop in Boston, May 29, 2008.
[08:08:06] <lairdp> localization and caches (ALTO), congestion (TANA)
[08:08:17] <lairdp> What's new/different about p2p?
[08:08:51] <victor.pascual> where could we find the slides? thanks
[08:08:53] <lairdp> client/server: target is a host, one or few IP's, so traffic optimization is finding the best network route to the destination.
[08:09:26] <jlcjohn> https://datatracker.ietf.org/meeting/72/materials.html
[08:09:45] <lairdp> http://alto.tilab.com/ [http://alto.tilab.com/] is the 'home page'.
[08:10:04] <lairdp> ALTO problem: peers have no network knowledge, leading to congestion.
[08:10:23] <lairdp> e.g. download data from tokyo when it could be retrieved from london (i.e. closer to dublin)
[08:10:34] xmlscott joins the room
[08:10:35] ray joins the room
[08:10:49] <lairdp> endpoints can't pick good peers - measurements are either poor or too expensive.
[08:11:30] <lairdp> for example, measuring throughput with thousands of peers would take a long time, and not tell you about anything other than throughput (e.g. reliability, cost)
[08:11:53] nov leaves the room
[08:11:59] <lairdp> propose defining an interface for a peer optimization service.
[08:12:35] <lairdp> e.g. "I am a peer and I have to exchange data that is real-time with anyone among X, Y, Z" and get back "Choose X"
[08:12:46] <lairdp> Or perhaps a cost model, etc.
[08:12:48] Lisa Dusseault joins the room
[08:13:29] <lairdp> ALTO doesn't care how topological data is provisioned, but does care how ALTO server talks with p2p networks (peers or servers)
[08:13:46] Henk Uijterwaal joins the room
[08:13:50] <lairdp> ALTO service providers are network providers, communities, or third parties.
[08:14:20] <lairdp> For example, Akamai knows the internet and could expose an ALTO server.
[08:14:40] <lairdp> Or an organization could operate on behalf of ISPs.
[08:15:08] <lairdp> Desired ALTO effect is to concentrate traffic within ISPs.
[08:15:11] victor.pascual leaves the room
[08:15:14] shikob leaves the room
[08:15:45] victor.pascual joins the room
[08:15:54] <lairdp> Gnutella simulations show localizing traffic improve performance and reduce transit.
[08:15:55] nov joins the room
[08:16:20] <lairdp> Also P4P field test results.
[08:16:35] DThaler joins the room
[08:16:39] <lairdp> Issues: topology hiding. ISPs consider network internls as reserved information.
[08:16:50] victor.pascual leaves the room
[08:16:56] <lairdp> Goal: provide network topology information without revealing network topology.
[08:16:59] CMW5S41 leaves the room
[08:17:11] gregory.juniper joins the room
[08:17:20] <lairdp> e.g. arbitrary priority valies (IDIPS), opaque ID's (P4P).
[08:17:39] Megumi joins the room
[08:18:05] <lairdp> Issue: Oracle location: is it centralized, etc.'
[08:18:20] <lairdp> Issue: Trust. What prevents misbehavior?
[08:18:41] <lairdp> For example, privacy violations, security issues, sub-optimal policies.
[08:18:43] shikob joins the room
[08:18:45] <lairdp> hint: ALTO is optional.
[08:19:01] <lairdp> Core Blocks:
[08:19:10] <lairdp> - discovery mechanism for finding the oracle.
[08:19:19] <lairdp> - request/response mechanism for querying oracle.
[08:19:29] victor.pascual joins the room
[08:19:42] <lairdp> Many use cases. Number of sources from 10 to 10^3.
[08:19:54] <lairdp> swarm file downloading.
[08:20:07] <lairdp> realtime communications (e.g. NAT relay for VOIP service).
[08:20:15] Henk Uijterwaal leaves the room
[08:20:17] <lairdp> p2p sreaming.
[08:20:22] Henk Uijterwaal joins the room
[08:20:53] <lairdp> mirror selection. e.g. finding the closest mirror via network topology as opposed to DNS.
[08:21:17] <lairdp> DHT's use proximity to maintain routing tables, using RTT's. Could use ALTO instead.
[08:21:37] <lairdp> Caching: are just a p2p peer, just very generous and fast.
[08:21:43] IP-CENTREX joins the room
[08:21:50] gaia joins the room
[08:21:54] <lairdp> But ALTO could be used to broadcast cache addresses.
[08:22:38] <lairdp> So should put some cache semantics into ALTO so that caches are (e.g.) sent as first peers to clients.
[08:22:52] rem joins the room
[08:22:53] sal joins the room
[08:24:00] <lairdp> cache location could use additional information (e.g. content ID, network ID) to be able to provide the correct cache.
[08:24:12] <lairdp> But networks may not want to disclose too much information.
[08:24:29] <rem> ping
[08:24:38] xiaohunhun leaves the room: Replaced by new connection
[08:24:41] ylee joins the room
[08:24:48] <lairdp> Question: is the problem reducing inter-domain traffic? Or is the problem more general?
[08:25:36] <lairdp> (asker is Stanislav)
[08:26:11] IP-CENTREX leaves the room: Replaced by new connection
[08:26:11] IP-CENTREX joins the room
[08:26:22] Henk Uijterwaal leaves the room
[08:26:30] kikuyuta joins the room
[08:26:36] Henk Uijterwaal joins the room
[08:26:37] <lairdp> Can already map IPs to ASNs.
[08:26:38] gregory.juniper leaves the room
[08:27:17] <lairdp> LookingGlass has inter-ASN transit data. Comment: it's a hack, not a protocol.
[08:28:18] <DThaler> Jon is hard to hear
[08:28:19] <lairdp> AD: the purpose of the BOF is to explore standardizing a protocol for acquiring this information.
[08:28:40] <lairdp> So even if the info is available now, a standard can make it easier/more accurate/easier.
[08:29:29] <lairdp> Stanslav: bit cost is an example of what an application cannot otherwise get.
[08:30:16] <lairdp> Bob Brisco: if the goal is reducing inter-domain traffic, what's next?
[08:30:58] lj324o230ar joins the room
[08:32:29] <lairdp> If internal uplink is congested, ISPs might prefer using available external links,
[08:32:31] xavi.mila leaves the room
[08:33:01] Megumi leaves the room: Computer went to sleep
[08:33:02] CMW5S41 joins the room
[08:33:12] <lairdp> If a query is specific to a particular network, or generic?
[08:33:19] lj324o230ar leaves the room
[08:33:20] kikuyuta leaves the room
[08:33:22] shikob leaves the room
[08:33:40] <lairdp> The goal is for the service to be generic to protocol and content.
[08:34:01] Megumi joins the room
[08:34:09] shikob joins the room
[08:34:27] <lairdp> The problem statement is very close, but needs to be "what is the optimal data source".
[08:34:56] <lairdp> For example, in wireless networks communicating with peers on the same cell tower might be bad due to resource contention in the tower.
[08:35:13] xavi.mila joins the room
[08:35:29] <lairdp> The answer might be to distribute routing "smarts" to the edges.
[08:35:48] spencerdawkins joins the room
[08:36:07] <lairdp> With that, the problem statement looks great. The problem statement is important, so we need to get it right before we get to charter bashing.
[08:36:24] shikob leaves the room: Replaced by new connection
[08:37:26] <lairdp> Answer: the goal is to allow ALTO servers to guide p2p apps to find "the best peer". The goal is not specifically inter-domain links. But the goal could be cost minimization, closest peers.
[08:37:29] jiangxingfeng leaves the room: Replaced by new connection
[08:37:36] Henk Uijterwaal leaves the room
[08:37:42] Henk Uijterwaal joins the room
[08:38:13] <lairdp> China Mobile: concerns about ALTO problem statement.
[08:38:26] Jack leaves the room: Replaced by new connection
[08:38:27] Jack joins the room
[08:39:06] gregory.juniper joins the room
[08:39:07] <Lisa Dusseault> I asked if the mics could be increased in volume a bit -- I guess they're working on it :)
[08:39:16] xavi.mila leaves the room
[08:39:21] shikob joins the room
[08:39:30] magnus joins the room
[08:40:07] <gregory.juniper> feedback i pretty wicked
[08:40:27] <Lisa Dusseault> yeah
[08:40:29] <gregory.juniper> on the mic's audio quality, that is (as opposed to the content of the comments ;-)
[08:41:04] <lairdp> Guidance needs to be driven by more than topology.
[08:41:27] nm leaves the room: Replaced by new connection
[08:41:28] nm joins the room
[08:42:06] shep joins the room
[08:42:17] <lairdp> Is 'I am peer P' claimed by the peer, or by a third party? Both, e.g. a BitTorrent tracker can ask about locality for a peer in the network.
[08:42:36] <lairdp> ALTO survey, Marco Tomsu.
[08:43:02] <lairdp> Note: I'm not getting all questioners' names. Can anyone fill in?
[08:43:44] Henk Uijterwaal leaves the room
[08:43:50] Henk Uijterwaal joins the room
[08:43:55] <lairdp> Guidance can be provided in many ways, organized as application level and 'layer cooperation'.
[08:44:13] <lairdp> - measure throughput with peers.
[08:44:39] <lairdp> (accurate, but limited to a small number of peers)
[08:45:14] enrico joins the room
[08:45:14] enrico leaves the room
[08:45:27] <lairdp> (used to select CODECs in Skype, peer data sources in BitTorrent, etc.)
[08:45:45] <Lisa Dusseault> Laird: feel free to interrupt somebody when they come to the mic to remind them to give their nme.
[08:46:25] <lairdp> IDMaps clusters IP addresses into ip prefix lists for each location, with 'virtual links' between locations.
[08:47:03] csp leaves the room
[08:47:05] <lairdp> Vivaldi introduces absolute network locations,
[08:47:49] <lairdp> (issue: triangular inequality is not always a given)
[08:48:03] <lairdp> (implemented as a plug-in into Azureus)
[08:48:15] xavi.mila joins the room
[08:49:12] <lairdp> iPlane is "an information plane for distributed services".
[08:49:15] jiangxingfeng joins the room
[08:49:44] <lairdp> builds a map that predicts routes between end nodes.
[08:50:32] <lairdp> ONO uses Akamai's topology info to optimize peer assignment.
[08:51:03] <lairdp> (uses akamai's DNS redirection to find peers that are near each other)
[08:51:19] spencerdawkins leaves the room
[08:51:43] <lairdp> (implemented as an azureus plugin)
[08:51:49] IP-CENTREX leaves the room
[08:51:58] xavi.mila leaves the room
[08:52:03] simon joins the room
[08:52:21] Henk Uijterwaal leaves the room
[08:52:27] Henk Uijterwaal joins the room
[08:52:36] <lairdp> P4P uses ISP-provided network maps that indicate distance, capabilities, business policies (e.g. time of day link utilization).
[08:53:01] <lairdp> (IPs are mapped into POP ID's for privacy and scalability)
[08:53:43] <lairdp> Oracle-based ISP-P2P Collaboration. Addresses are ranked by an "oracle" that knows ASN, BGP, etc.
[08:54:17] bless joins the room
[08:54:21] xavi.mila joins the room
[08:54:22] <lairdp> IDIPS: ISP Driven Informed Path Selection uses guidance to optimize for multiple sources and multiple destinations.
[08:54:33] <lairdp> (presented at IETF 71)
[08:56:22] <lairdp> Question: Most ISPs don't own their own facilities, but rent facilities from other ISPs. Many ISPs operate in many countries. ALTO concept is broken because ISPs can't know their topology.
[08:57:15] <lairdp> Jon: This is a survey of the approaches, not a proposal to standardize on any of these approaches.
[08:57:24] psavola leaves the room
[08:58:23] CMW5S41 leaves the room: Replaced by new connection
[08:58:24] lairdp leaves the room
[08:58:24] CMW5S41 joins the room
[08:58:32] <bless> anyone else having problems with the audio stream (not really audible, obviously too many packet drops and stalled connections)
[08:58:35] <bless> ??
[08:58:39] gregory.juniper leaves the room
[08:58:55] <jlcjohn> audio was OK earlier
[08:59:05] <jlcjohn> pretty useless now
[08:59:31] <jlcjohn> problem doesn't seem to be between me and uoregon
[09:00:05] <jlcjohn> (i.e. no packets dropped from uoregon
[09:00:44] <sebastian.kiesel> audio was very bad in the room - someone has modified the mike settings. now it's better in here, but maybe that screwed network audio
[09:02:09] <sebastian.kiesel> henning now talks about netwok costs
[09:02:35] <sebastian.kiesel> a large bandwidth problem is coming. goal is to minimize cost of delivery.
[09:02:39] nm leaves the room: Replaced by new connection
[09:02:39] nm joins the room
[09:02:41] <Lisa Dusseault> For anybody who didn't know, Marco's slides were already online
[09:02:41] <Lisa Dusseault> http://www3.ietf.org/proceedings/08jul/slides/alto-3.pdf
[09:02:58] <Lisa Dusseault> generally: https://datatracker.ietf.org/meeting/72/materials.html
[09:03:03] <sebastian.kiesel> (this is laird, continuing notes)
[09:03:22] <sebastian.kiesel> bandwidth cost is a commodity. inter-domain costs are well understood.
[09:03:45] <sebastian.kiesel> roughly 10 cents/gb.
[09:04:18] <sebastian.kiesel> so there's an economics problem - a DVD costs $1 to transport in pure bandwidth cost.
[09:04:56] <sebastian.kiesel> cost has a step function - bandwidth in the home is cheap, data outside ISP is expensive, etc.
[09:05:58] <sebastian.kiesel> goal is to sort economics of data sources, not to precisely predict transfer times.
[09:06:44] <sebastian.kiesel> isps may choose to make economics explicit instead of implicit.
[09:07:07] Lars joins the room
[09:07:10] <sebastian.kiesel> the question: do we need query protocols to allow isps to expose costs of data transfers?
[09:08:10] <sebastian.kiesel> upstream and downstream may differ.
[09:08:18] kazu leaves the room: Replaced by new connection
[09:08:34] <sebastian.kiesel> applications can then make rational decisions.
[09:09:30] <sebastian.kiesel> Henry: Verizon diagram accurate. But backhual costs are the 'killer' costs.
[09:11:02] mhp joins the room
[09:11:05] <sebastian.kiesel> back-hual is the connection from the provider to the internet.
[09:11:14] nm leaves the room: Replaced by new connection
[09:11:15] nm joins the room
[09:12:17] <sebastian.kiesel> or connection from access network to core network.
[09:13:27] <sebastian.kiesel> david brian: price and bandwidth issues are different for different isps.
[09:13:40] <sebastian.kiesel> it's important to expose info to users as well as isps.
[09:14:04] <sebastian.kiesel> answer: goal is to allow all participants to make rational economic decisions.
[09:15:57] <sebastian.kiesel> bob brisco: the only cost is the 95th percentive - bandwidth is free except for congestion.
[09:16:45] <sebastian.kiesel> rich woundy, comcast. the predominant cost is the access network, not core.
[09:17:49] <sebastian.kiesel> don't want to surprise users with $1000 bills.
[09:18:00] <sebastian.kiesel> answer: industry is heading towards volume based pricing.
[09:18:11] <sebastian.kiesel> so either surcharging or capping, neither are good.
[09:19:18] <sebastian.kiesel> ron desliva, time warner cable.
[09:19:27] <sebastian.kiesel> don't want to expose internal costs, just external costs.
[09:20:31] <sebastian.kiesel> ? time shifting all transfers to night will make night time the peak, making daytime available.
[09:20:57] <sebastian.kiesel> (IMO, this is an example of how things can work well - exposing economics can change application and user behavior to use network more efficiently)
[09:21:03] <sebastian.kiesel> Sebastial, ALTO requirements.
[09:21:32] <sebastian.kiesel> P4P and NAPA-WINE are related projects.
[09:21:58] Henk Uijterwaal leaves the room
[09:22:05] Henk Uijterwaal joins the room
[09:22:13] <sebastian.kiesel> example figure.
[09:22:21] leslie-ietf joins the room
[09:22:32] enrico joins the room
[09:22:44] <sebastian.kiesel> the alto solution should be more general than this example, but...
[09:23:27] <enrico> it seems we need a scribe
[09:23:43] <enrico> anyone in the room could please relay what's been said?
[09:23:51] <sebastian.kiesel> Walk through example...
[09:24:10] <sebastian.kiesel> (this is Laird) I've been scribing. Aren't the messages getting through?
[09:24:27] xavi.mila leaves the room
[09:25:03] <sebastian.kiesel> ALTO should apply to may applications, not just p2p swarm file delivery. e.g. finding VOIP relay.
[09:25:04] <saverio.niccolini> they are getting through as Sebastian.Kiesel was writing (at least they apear in my client)
[09:25:05] magnus leaves the room
[09:25:07] shikob leaves the room
[09:25:20] Lars leaves the room
[09:25:20] bless leaves the room
[09:25:36] <sebastian.kiesel> (That's it - my client got thrown off of jabber server, so I'm using Sebastian's laptop - Laird)
[09:26:02] <sebastian.kiesel> Feeding topology into device is implementaiton specific, could be BGP, static configuration, etc.
[09:26:02] <saverio.niccolini> ok :-)
[09:26:16] jiangxingfeng leaves the room: Replaced by new connection
[09:26:19] <sebastian.kiesel> Coordination between ALTO servers should be deferred.
[09:26:32] <sebastian.kiesel> So focus on ALTO clients querying ALTO servers.
[09:26:55] <sebastian.kiesel> Need ALTO server discovery.
[09:27:17] josoinin joins the room
[09:27:25] <sebastian.kiesel> ALTO might rate IP addresses, or network locatoin (e.g. ASN, IP prefix).
[09:27:31] enrico leaves the room
[09:27:40] enrico joins the room
[09:27:53] richard.barnes@gmail.com joins the room
[09:28:05] Henk Uijterwaal leaves the room
[09:28:12] Henk Uijterwaal joins the room
[09:28:30] <sebastian.kiesel> rating should be on topology, business rules. rating should not be instantaneous network conditions.
[09:28:42] <sebastian.kiesel> ALTO does not replace congestion control.
[09:29:29] simon leaves the room
[09:29:37] <sebastian.kiesel> question: does ALTO communicate the 'decay rate' of the value of the information? 
[09:30:06] <sebastian.kiesel> answer: there's a 'lifetime'. Followup: decay rate is better than lifetime, as the value of the data doesn't expire at a time.
[09:31:18] simon joins the room
[09:31:31] <sebastian.kiesel> question: ALTO is proposing a specific way to do things, and are we sure that we're doing it the right way?
[09:31:44] <sebastian.kiesel> answer: considering alternatives. e.g. name of resource rather than list of addresses.
[09:32:08] Andrew McGregor joins the room
[09:32:21] <sebastian.kiesel> Alan: we're focusing on source selection, but p2p is bi-directional and peers also select data destinations.l
[09:32:42] <sebastian.kiesel> henry: long list of contributors.
[09:32:58] <sebastian.kiesel> question: shouldn't p2p providers be at the table?
[09:33:33] <saverio.niccolini> P4P folks are in the requirements draft
[09:34:01] tmonca joins the room
[09:34:36] <sebastian.kiesel> answer: Pando was a co-author, BitTorrent and others were involved in the discussions.
[09:34:37] bless joins the room
[09:34:40] <richard.barnes@gmail.com> saverio, you mean they're authors on the requirements draft?
[09:35:30] <sebastian.kiesel> (This is Laird Popkin, using Sebastian's jabber client) Pando is a p2p network and was a co-author of the requirements.
[09:35:41] <saverio.niccolini> right, this was my point
[09:35:54] <sebastian.kiesel> As the co-chair of the P4P Working Group, I (somewhat) represent a large number of p2p companies.
[09:36:18] <sebastian.kiesel> Also, BitTorrent people were very involved in the discussions, though not as a co-author of the requirements document.
[09:36:43] Henk Uijterwaal leaves the room
[09:36:46] <sebastian.kiesel> But (to editorialize) it'd be good for more p2p providers to be involved in ALTO....
[09:36:59] richard.barnes@gmail.com leaves the room
[09:37:12] <saverio.niccolini> I agree
[09:37:50] <sebastian.kiesel> security and privacy issues. the ALTO server may want to tell the clients not to send content IDs because they don't want to know.
[09:37:53] victor.pascual leaves the room: Replaced by new connection
[09:38:04] <saverio.niccolini> anyway vendors (e.g. NEC, Cisco) are not willing to kill P2P if they can sell equipment making money out of it, jsut my personal view
[09:38:49] Jan joins the room
[09:39:04] <sebastian.kiesel> open issues: is ALTO a service or a file format?
[09:39:40] <sebastian.kiesel> should cache and peer discovery be one mechanism or two?
[09:39:57] <sebastian.kiesel> how to avoid privacy issues?
[09:40:10] <sebastian.kiesel> do we need special cache handling attributes, or are they just peers?
[09:40:23] <sebastian.kiesel> can ALTO be generalized for non-p2p apps that need a similar service?
[09:40:25] Alissa Cooper leaves the room
[09:41:14] xmlscott leaves the room
[09:41:38] <sebastian.kiesel> question: operators may not want to know what content is, but if cached may want to.
[09:42:17] <sebastian.kiesel> Renaldo presenting Caching and Peer Selection in ALTO.
[09:42:42] <sebastian.kiesel> focusing on cache discovery, not caching strategies
[09:42:53] <sebastian.kiesel> also alto server discovery, might be closely related.
[09:43:29] Lars joins the room
[09:43:55] <sebastian.kiesel> peers may bootstrap using DNS SRV, DNS-BP, BEP-22, etc.
[09:44:14] <sebastian.kiesel> Or peers contact ALTO server and ALTO server provides cache locations.
[09:44:37] <sebastian.kiesel> And either of the above can be used to find other resources (e.g. STUN servers).
[09:44:59] nov leaves the room
[09:45:28] <sebastian.kiesel> bittorent: peers request peer list from tracker, then peers exchange data.
[09:46:01] <sebastian.kiesel> BEP-22 has ISP operating its own tracker that returns peers within the ISP.
[09:47:12] <sebastian.kiesel> next step: discovery document for alto server and caches.
[09:47:20] Henk joins the room
[09:47:36] gregory.juniper joins the room
[09:48:55] <sebastian.kiesel> Is there a 'chasm' between discovery new resources (e.g. caches) and rating of existing resources.
[09:49:01] ray leaves the room: Replaced by new connection
[09:49:02] enrico leaves the room
[09:49:28] Arifumi joins the room
[09:50:06] Alex McMahon joins the room
[09:50:07] <gregory.juniper> Dave O: do you have a suggestion for an existing discovery proto that would be a good fit?
[09:50:32] ray joins the room
[09:50:41] <gregory.juniper> Looks like Dave O isn't in the jabber room
[09:50:49] <gregory.juniper> others want to answer that then?
[09:50:50] <sebastian.kiesel> Alan: should support both modes.
[09:51:24] <sebastian.kiesel> Similar Problems: 
[09:51:54] <sebastian.kiesel> IP4 vs. IP6 dual stack problem. Decide whether it's better to use IPv4 or IPv6 routes.
[09:52:34] <sebastian.kiesel> multi-homing problem choosing between multiple paths.
[09:52:50] <sebastian.kiesel> server replicas: selecting between available mirror servers
[09:53:02] Lars leaves the room
[09:53:35] <sebastian.kiesel> ALTO best peer selection.
[09:53:51] <sebastian.kiesel> How to select best peers from the swarm.
[09:54:06] <sebastian.kiesel> similar problem, but in a p2p application context.
[09:54:40] Henk leaves the room
[09:54:46] Henk joins the room
[09:55:23] <sebastian.kiesel> ALTO can be used to solve this common problem if
[09:55:34] <sebastian.kiesel> - transport addresses include both IPv4 and IPv6.
[09:55:44] <sebastian.kiesel> ALTO isn't restricted to p2p.
[09:55:55] <sebastian.kiesel> Charter Discussion
[09:55:58] enrico joins the room
[09:56:02] Lars joins the room
[09:56:11] gregory.juniper leaves the room
[09:56:37] Jan leaves the room: Replaced by new connection
[09:56:37] jmgm joins the room
[09:57:35] <sebastian.kiesel> questioners: please state your names!
[09:57:46] ldondeti joins the room
[09:58:26] <sebastian.kiesel> proposing wg to produce prooootocol to enable communications between p2p systems (and others) and informed external services.
[09:58:38] <sebastian.kiesel> Needs to address tracker-based, trackerless, structured, unstructured, etc.
[09:59:38] <sebastian.kiesel> protocoL
[09:59:42] <sebastian.kiesel> - discovery of oracle
[09:59:48] <sebastian.kiesel> - query/response with oracle
[10:00:44] <sebastian.kiesel> work items: discovery of 'special' nodes such as caches, media relays, etc.
[10:00:47] Henk leaves the room
[10:00:54] Henk joins the room
[10:01:13] <sebastian.kiesel> in-line means that the oracle sends new appropriate nodes to applications.
[10:01:31] Alex McMahon leaves the room: Computer went to sleep
[10:01:55] <sebastian.kiesel> off-line means the application would send a list of possible special nodes to the oracle and receive a recommendation.
[10:02:55] jmgm leaves the room
[10:03:02] spencerdawkins joins the room
[10:03:50] rem leaves the room
[10:05:23] <sebastian.kiesel> is alto (ted) trying to solve the general resource discovery problem, or just alto server discovery.
[10:06:50] <sebastian.kiesel> separation is a good thing
[10:07:55] <sebastian.kiesel> had a similar set of discussions for speech fc.
[10:08:02] spencerdawkins leaves the room
[10:08:05] josoinin leaves the room
[10:08:11] <sebastian.kiesel> requirements and preconditions must be documented.
[10:08:18] <sebastian.kiesel> requirements are what we're trying to solve.
[10:08:34] <sebastian.kiesel> a precondition could be that we're relying on an existing mechanism for finding ALTO servers.
[10:08:50] lairdp joins the room
[10:09:02] <lairdp> I'm back as myself.
[10:09:32] CMW5S41 leaves the room: Replaced by new connection
[10:09:33] CMW5S41 joins the room
[10:10:15] josoinin joins the room
[10:10:20] <lairdp> Jon: we're going to have a requirements phase, that should clarify this. We may not need to address all of this now.
[10:10:28] <lairdp> But it should be soon.
[10:10:41] <lairdp> But to clarify: ALTO won't reinvent service discovery.
[10:11:07] <lairdp> The question is whether there's a consensus that the IETF wants to take on building a service like this.
[10:11:11] tmonca leaves the room
[10:11:30] <lairdp> Can we identify a discrete scope? Is this what we want to do?
[10:11:55] Henk leaves the room
[10:12:01] Henk joins the room
[10:12:36] <lairdp> Microsoft: the discovery mechanism is a distraction. The peer selection is the real issue. Is there a definition of what "best" means?
[10:13:13] <lairdp> Gregory: Is the charter mandating that an existing discovery protocol be used?
[10:13:57] <lairdp> Jon: Not mandating, but it seems profoundly unlikely that ALTO would define a new discovery protocol, just that the requirements for a discovery protocol be defined.
[10:18:03] nm leaves the room: Replaced by new connection
[10:18:04] nm joins the room
[10:18:07] xavi.mila joins the room
[10:18:40] ray leaves the room: Replaced by new connection
[10:18:40] ray joins the room
[10:18:45] <lairdp> IBM: overlaps P2P SIP group. Has there been a division or coordination between the two groups?
[10:19:15] <lairdp> Talking with ADs and chairs, P2PSIP could use ALTO.
[10:19:24] Jan.Seedorf leaves the room: Replaced by new connection
[10:19:25] josoinin leaves the room
[10:19:31] ray leaves the room
[10:19:38] ray joins the room
[10:19:46] <lairdp> But given the numbers, optimizing p2p file transfers might be worth doing even without P2PSIP.
[10:21:15] <lairdp> Question: many items are described as 'bittorrent does X' and others are 'bittorrent could be done by...'. We should review everything to make sure that we're not solving problems that are already solved.
[10:21:28] <lairdp> Scope:
[10:21:58] <lairdp> focus on interface. algorithms for deciding what to recommend are not in ALTO.
[10:22:26] <lairdp> ASNs, coordinates, priority are in scope. Congestion is out of scope.
[10:22:35] <lairdp> Issues related to content (e.g. legality) are not in scope.
[10:24:03] Jack leaves the room
[10:24:03] <lairdp> Question:
[10:24:11] <lairdp> - Need to figure out who to ask (discovery)
[10:24:29] <lairdp> - Here's a list, give me back a recommendation for best source to me.
[10:24:33] <lairdp> - Give me costs.
[10:25:19] brabson leaves the room: Replaced by new connection
[10:25:19] brabson joins the room
[10:25:23] <lairdp> Lars: TANA is about congestion control and data transmissions. ALTO is about asking other sorts of questions.
[10:25:35] <lairdp> ALTO should ask about slow-changing things, not fast-changing things.
[10:25:40] nm leaves the room: Replaced by new connection
[10:25:40] nm joins the room
[10:26:01] shikob joins the room
[10:27:52] <lairdp> If an app asks ALTO for things like current RTT and jitter on link, that's a problem.
[10:28:25] <lairdp> Comment: the protocol part is easy, but the semantics are complex.
[10:28:44] <lairdp> A requirements document is how we typically work through issues, but that feels like a poor way to tackle this.
[10:29:42] enrico leaves the room
[10:29:54] ttfr joins the room
[10:30:02] <lairdp> Vijay: is this a problem that the IETF should solve?
[10:31:23] brabson leaves the room
[10:31:28] shikob leaves the room
[10:31:48] ttfr leaves the room
[10:31:48] lairdp leaves the room
[10:33:17] lairdp joins the room
[10:33:38] <lairdp> hum 'pro' and 'con' were comparable. the 'con' were quite vehement.
[10:33:55] Megumi leaves the room: Computer went to sleep
[10:34:21] <lairdp> taking a break to ask a question. can someone else scribe for the interim?
[10:36:50] leslie-ietf leaves the room
[10:37:32] shikob joins the room
[10:38:34] <lairdp> One concern: is this too researchy?
[10:38:39] Lars leaves the room
[10:38:40] ylee leaves the room
[10:38:41] shikob leaves the room
[10:38:50] nm leaves the room
[10:39:07] shikob joins the room
[10:39:10] <lairdp> The concern is 'is the 'oracle' model the right approach, or are we making an solution architecture decision prematurely.
[10:39:36] shikob leaves the room
[10:39:38] jinhong leaves the room: Computer went to sleep
[10:40:03] CMW5S41 leaves the room
[10:41:13] ldondeti leaves the room
[10:42:13] Lisa Dusseault leaves the room
[10:42:19] gaia leaves the room: Computer went to sleep
[10:42:25] Damon.Wischik leaves the room
[10:42:31] Henk leaves the room
[10:42:34] xavi.mila leaves the room
[10:42:47] saverio.niccolini leaves the room
[10:42:58] Andrew McGregor leaves the room
[10:44:10] mhp leaves the room
[10:44:46] lairdp leaves the room
[10:49:39] shep leaves the room: Logged out
[10:50:22] sebastian.kiesel leaves the room: Replaced by new connection
[10:55:42] DThaler leaves the room
[10:58:04] Dan York joins the room
[10:58:11] Dan York leaves the room
[10:58:32] ray leaves the room
[11:00:26] Lars joins the room
[11:06:06] Haurhiko Nishida leaves the room
[11:22:35] nov joins the room
[11:24:11] nov leaves the room
[11:30:37] DThaler joins the room
[11:30:45] Lars leaves the room
[11:31:47] kikuyuta joins the room
[11:31:47] kikuyuta leaves the room
[11:51:47] DThaler leaves the room
[11:58:01] Arifumi leaves the room
[11:58:11] simon leaves the room
[12:07:40] amalis joins the room
[12:07:58] amalis leaves the room
[12:15:51] Alex McMahon joins the room
[13:29:59] Alex McMahon leaves the room
[14:05:09] bless leaves the room
[15:05:10] sebastian.kiesel joins the room
[15:05:46] sebastian.kiesel leaves the room
[21:42:41] jlcjohn leaves the room