notes - IETF71 Thursday Plenary notes

Thursday, 13 March

IAB/Technical Plenary Agenda:

1. Welcome by Olaf Kolkman


2. IRTF Chair's Report by Aaron Falk

Aaron:  Seven research groups met this week.  A few are meeting tomorrow.  If you are a working group that is spending time on hard problems in the areas of the RGs, bring them up at the RG meeting.  It might be something that the RG is working on.  We closed the end-middle-end research group and internet measurement group.  The P2P research group is lingering, and we're hopeful to find capable chairs that are in touch with the research issues.

Aaron:  The IAB met with the NMRG this morning.  We are still working on the IRTF RFCs - three IDs are in the editor's queue and a few are in the approval process.  The IAB is interested in work on unwanted network traffic.  There was also a bar BOF this week on network virtualization.  There was good attendance of about 20.  It is not a RG yet, but there is some interest.  There is also interest in QOS policy framework.  Once this area is a little better understood, we may start an RG.

Aaron:  The anti-spam RG is back!  They are interested in finishing two drafts, which should be revised soon.  There is also a wiki for the RG that attempts to catalogue spam mitigation techniques, and why some shouldn't be used.  The RRG has a lot of energy.  They are still getting new proposals, with the plan to develop a consensus by next March.  The RRG chairs provided a clarification this week of the near term intent to look at the architectural issues, above the details of specific mechanisms.  

Aaron:  The DTNRG has been very active.  Implementations are showing up all over the place.  There are a huge number of drafts and published RFCs.  There was a discussion in the meeting of possibly moving the reference code to sourceforge, making it more of a community activity.  The HIP RG is about to publish a draft on NAT / firewall traversal issues.  They are looking to extend the functionality of HIP, and there is interest by P2P SIP.  

Aaron:  The ICCRG has taken on the role of reviewing congestion control proposals before they reach the transport area.  Compound TCP, for example.  This is now available in MS Vista (though it is not on by default), and there is also a LINUX implementation.

Aaron:  The NMRG is trying to learn about the performance issues of NM protocols, spending time figuring out trace formats, aggregation of data, etc.  We had an IAB review this morning, and they were encouraged to get a little closer to NANOG and RIPE.  They have developed a list of topics as a research agenda.  For those in the operator community running into research issues, contact Juergen Schoenwaelder.

Aaron:  SAMRG is looking at a multicast framework, where you have multiple kinds of multicast, and how to integrate them into a common framework.  They are currently prototyping, and are planning a meeting at MILCOM.  MOBOPTs has been active in the areas of privacy and IPv6, and are finishing up a problem statement on multicast mobility.



3. IAB Chair's Report by Olaf Kolkman

Olaf: We have had a change in IAB membership.  (see RFC 2850)

Fred Baker thanks the outgoing IAB members. They all get a plaque.

Olaf:  The IAB has a few documents in process.  We re close to finishing documents on "Principals of Internet Host Configuration", and "What Makes for a Successful Protocol".  The publication of "DNS Choices" is pending.  What does this mean?  The IAB is intending to publish it.  Referring to 4845, this describes a method to solicit feedback from the community.  It is not a last call; IAB documents are IAB consensus, but input from the community is taken seriously.  We will be more clear in the future.  

Olaf:  A large part of our work goes into admin issues, and inter-organizational relations.  In the last few months, we responded to a request for information by the U.S. Department of Commerce.  (about private sector hand-off).  With ICANN, there was an open call for feedback and comments on the stability of DNS while adding new GTLSs.  We provided reference to the RFC, and suggested review by ICANN on a per GTLD basis.  

Olaf:  The IAB has been working with the ITU-T on T-MPLS.  We have reported on this before.  At the last IETF we established an ad-hoc group to work with Study Group 13, and since then have done a ton of work to ensure good outcomes for both organizations.  This coordination was a success, and the ad-hoc group has concluded.  Now there is a joint project team established - this is new for the IETF.  It is a group of about 20 specialists, an ITU-T and MPLS interoperability design team sponsored by the routing area.  The work will be done in the IETF.  It was set up to assist the ITU in making a choice from two options:  move T-MPLS into the IETF (taking into account mutual requirements), or establish a clean separation of name, eth type, etc.  An enormous amount of work went into making this happen.  Special thanks to Dave Ward, Ross Callon, Monique Moreau (?), Stuart (Cheshire?), and others.  And now, Barry Leiba will introduce the technical topic.



4. Technical topic: IPTV
   Introduced by Barry Leiba

Barry:  When working on this topic, I ran into this story in the New York Times.  It talks about TIVO and youtube.  One aspect of what's going to be talked about - using a protocol stack that we've developed here to provide video over a private network.  

Barry:  We have two speakers... (introduces Marshall Eubanks and Keith Ross).  



5. The Video Tsunami: Internet Television, IPTV and the coming wave of Video on the Internet
   Marshall Eubanks, AmericaFree.TV



Fred: the point of the article in the newspaper was that lack of investment in the US could effect business in the US.

Marshall: clearly there is a lot more relative investment in broadband elsewhere (in Asia and Europe). The US runs the risk of falling behind - innovations are occurring there.

Henning Schulzrinne: comment regarding the difference between the channel and publisher

Marshall: it really is about push vs. pull. Both have a role.  Very large contents are a pull thing, you pull that to you and maybe want to watch it later.  But sometimes you might not know that you want to see something, that's why companies will increasingly follow the push model. But switching on the TV and just watch what's on, will not go away.



6. Peer-to-Peer Internet Video
   Professor Keith Ross, Polytechnic University, Brooklyn



Aaron Falk: why do you think these P2P services show up in China mostly, as opposed to other places?

Keith Ross: There are a number of start-ups in the US trying to do the same thing - i.e. white-lbel P2P services. In China the TV stations are run by the government. They have allowed for those types of P2P systems - they are not illegal.

Cullen Jennings: interested in research data with the 'tit for tat'
algorithms.  If the algorithm is sending more yields better service, the optimal strategy is to fully saturate the network.  Is there any work on that? People who are investing in this. How are they dealing with saturation issues?

Keith: not aware of any research in that area.
BitTorrent gets best performance at highest throughput.

Cullen:  I am not driving at the congestion point.  But should we expect this saturation behavior in the future?

Keith:  Even with tit-for-tat, you reach a max.  Once you go beyond that...there is a limit to how good the service can be.


Gentleman from China: regarding the question why there is so much development in China: one reason is that there are so many local TV stations and channels it is unbelievable.  The channels are not centralized - that encourages people to distribute it more.

Eric Rescorla:  As I understand it, the premise of most of these approaches is to prevent freeloading.  is there research to detect cheating in this area?  From the state of the research it looks very difficult to prevent colluders.

Keith: Recent research is showing that you can be a free rider and still download files at high rates if you manipulate the protocol. The research shows that it can be done, but in practice it is not really done. And there is no real way around this. If people hack the protocols and the clients, that can be done.

Eric Rescorla:  If everyone knew how to modify the protocols, would people cheat?  But people don't know how.  How important is an enforcement mechanism, or is it enough to just have the clients be reasonable like TCP?  They are not guaranteed to work, but they usually do.

Keith: yes, there is a mechanism but it is not guaranteed to work. However, in practice it works pretty well (just like IP).

Pasi Eronen: How relevant will be real-time streaming vs. faster-than-real-time downloading?  Storage is getting cheaper faster than bandwidth.  If you beloeve that DRM will go away (not economically viable), and you can do mash-ups, you become an important part of the ecosystem.  E.g. replace a character in a movie with an avatar of yourself..

Marshall: Nowadays a Terrabyte is not that much Multicast distribution.  I think networks will get so fast that these will look like minor differences. 

Dave Crocker: reference to older model of dealing with burstiness: seems to be a fundamentally different type of traffic than the Internet was originally built for. Seems that for continuous streaming you ought to have different handling even if it is all based on IP.  Steady flow of data (the Internet not designed for that)  vs. pieces of data.

Marshall: no, we don't really see that. People don't normally watch for that long.

Dave Crocker:  I think we could see more of that - perfectly reasonable to leave a stream on in the background all the time.

Howard Green: regarding importance of delay: How big a delay or lag is acceptable?  Most US television has a delay deliberately inserted.  It seems a characteristic of most P2P systems is an increase in latency.  Do we believe that people pay extra for TV where the latency is significantly smaller?

Keith: a more serious issue is channel switching lag. That will be a big challenge for future research.

Marchall: that is specifically important in the gambling area.  Day-traders are down to 10s of ms.  Maybe news of an airline crash - 5 sec - get the stock trade in.

Philip Hallam-Baker: didn't quite like either approach: in P2P he doesn't like the idea that all the data moves up and down again. Couldn't the providers not provide caches? Then it doesn't have to be peer-2-peer. Also doesn't like proposals that change the Internet infrastructure (like multicast). It needs an economic model. Once you have an incentive for the ISP to deploy it for their own value, they will also do it for the user.

Marshall: typical cable plant has maybe 1000 channels. And there may be 100 channels with only 1 watcher at a time. Then multicast would be really good.

James McEachern: interested to know more about the interdiction efforts. Has it been tried to stop P2P by attacking it? Is there any research on that?

Keith: fair amount of research with interdiction companies. Rather successful in attacking. Attacks were much less successful for BitTorrent. Not aware of attack on Nettraversal. He is aware of the following: Fake torrents, fake blog attack, connection attacks.

Charlie Perkins: question about 'tit for tat' for video updloads?  My understanding is that with audio uploads, you don't get credit for uploading something that everyone else uploads, but rather uploading something more rare gives you more credits.  But in this video model, how can you distinguish what's valuable? You could just upload useless stuff (but volume) all day.  How could you check against this?   

Keith: 'tit for tat' algorithm works between two trading parties: we want the same video and I have chunks you want and you have chunks I want.

Marshall: how do you start getting credits in 'tit for tat'?  Are you aware of research that shows if the calibration of all that is useful?

Keith: there is some research about incentives in p2p systems. In fact there is only one that is successful and that is the one implemented in BitTorrent. That is the algorithm described above: I only trade with you if you trade with me.  That is the only credit you get 0 there is no trust required because it is immediate.

Marshall: heard about systems where you can build up credits. That would mean an whole new economy would be built.

Keith: yes, heard about that something like this has been proposed, but is not aware of any system that has implemented that successfully.


7. Open Microphone Session (old and new IAB on stage)

Philip Hallam-Baker: very impressed by the IPv6 experiment. Only problem he has: the results were expected. It was a demonstration. The structural problem of moving from IPv4 to IPv6 was exposed. The fact that this is a deployment problem for IPv6. Why wasn't the IAB predicting this 6 - 8 years ago when it might have been more useful. What is the institutional problem? How are we going to deploy changes to the Internet infrastructure? It is not impossible.

Leslie: In fact, the IAB was signaling it as important 6 - 8 years ago, but it was difficult to get attention focused.  However, there are a number of differences between then and now. One difference is that now we have enough criticial mass and there are enough pieces, equipment, software out there that can be tested. We could have talked about it in the past only hypothetically, now we can get our hands dirty.  It's worth noting that there are issues on the table now that the IAB is pointing out for which there is a similar problem of trying to talk about issues in advance of readiness for deployment.  We need to pay attention.

Olaf: the IAB is struggling with this issue (see the IAB document on successful protocols).  We don't have the perfect crystal ball.

Elwyn: was looking at how we might help IPv6 working at the time. Difference between then and now: an awful lot is about application organisation. It is really about things like web references etc. That's not the base engineering stuff, it is mostly about deployment. Not really about the network architecture. About the IPv6 experiment here at the IETF: the fact that we got all these APs and other equipment was great and would not have been possible  6 - 8 years ago.

Aaron: believes that it was a good demonstration. Because it wasn't very exciting. It is good that it wasn't more exciting.

Philip: we were not successful in showing that IPv6 is ready to be deployable. It took him a long time to configure everything on his laptop.

Olaf: many people have not played with IPv6. It is not part of our nature to go home and play with it and 'eat our own dog food'.  That is actually a scary thought.

Lixia: there are many differences between then and now: the limitations of NAT was not as obvious, there were not as many p2p applications and IP wasn't as big as today. Secondly: pressure was not as big (i.e. IPv4 address space was not as scarce).

Joel Jaeggli: is concerned that there is not enough operational experience on the IAB. It is completely terrifying that people who run large networks don't participate in the standards process. How do we get them into the WGs and see their needs represented by the people who are on the IAB. Hopes that those on the IAB who do have that experience are fighting very strongly for their needs. More outreach is needed.

Kurtis: there are many people here with operational experience. But people are only concerned with their own little network and not with the overall structure anymore.

Gregory Lebovitz: we are going to spend time with a number of operators in the next few weeks to do a survey to find out more about their needs. Another point is that enterprise networks have grown significantly (together with their traffic). That has changed the economics. Hopes we can reach out to those operators as well and understand their requirements as well.

Andrew Malis: he has an architectural position at Verizon, but he has a lot to do with the operational people. In the past he has chaired IETF ops WGs. Before that he helped to operate the ARPA net.

Danny McPherson: a lot of the onus is on the operators to show up and speak up. He does not believe that the operators are under-represented more than any other group.

Loa Andersson: Olaf talked about the interaction with the ITU. Operators commented on that. While he shares the concern about the absence of operators in general, they seem to be aware of what is going on and step in when necessary.

Iljitsch van Beijnum: the IPv6 was a good experiment and showed good results. What are the next steps?

Leslie: wonders whether having an IPv6 only network at the IETF meetings in the future would encourage people and play with their software and protocols. Would that be useful?

Gregory: yes, maybe we can extend those IPv4 outages and get operators and vendors to work on this.

Russ: the IETF meeting survey will contain a question: did you find the experiment useful and shall we repeat it

Olaf: asks the audience: who would like to have an IPv4 outage during the plenary session at the next IETF?
- many hands up (roughly half)

And who is opposed to that?
- very few hands up

Who would like to have an IPv6 only SSID during the week? 
- many hands up (roughly 2/3)

And who is opposed?
- no hands up

Mark Andrews: thanks google for jumping in make an IPv6 web site available. He would like to encourage other vendors to do the same (how can you sell IPv6 equipment without an IPv6 presence on the Net?). Encourages everyone to go home and set up a tunnel and test IPv6.

Dave Crocker: the focus of the experiment is to take current work and to check if it is useful.  Something like the DNS choices document is an attempt by the IAB to facilitate the thinking about strategic decisions. These are the sorts of strategic things the IAB should be addressing. The IESG cannot do strategic stuff, because they are busy with daily stuff.  The IAB is the only place where one can take a step back and where strategic stuff can happen. Encourages the IAB to not get drawn into day-to day issues.

He further suggests to include text in each IAB document describing what the intend is of that document and what the community is supposed to do with IAB advice.

Olaf: very hard to come up with a message that offers contemplation and trade offs.

Gregory: when the IAB publishes something, people can read and then at the next IETF it can be discussed by those who are interested in it.

Dave Crocker: good idea

Bob Briscoe: there is a cultural divide and it does need both viewpoints (standards and operators). Feels that this clearly is not an operators forum.  Another point: Do you think BitTorrent vs. the rest of the Internet users is a problem?

Dave Oran: we really need to look at the problem that p2p applications will find new bandwidth and they will suck it up. Payback for new bandwidth is much less clear. Proposes the IAB look at that (operational and architectural). Otherwise it could happen that providers block applications vs. users trying to work around that.

Kurtis: blocking p2p traffic seems to assume that the users don't intend to send that traffic.

Bob: it is about congestion, p2p just happens to be the type of traffic that uses a lot of bandwidth.

Kurtis: true, that is a different problem. 

Andrew Sullivan: regarding the operators issue: understands that there is a problem with operators not showing up. On the other hand is that not also a measure of success that operators don't show up anymore. Maybe there is nothing for them to do here?

Kurtis: the absence of operators can also be interpreted differently: maybe the operators were fed up that they couldn't get anything done.

Andrew Malis: he is here because operators still have requirements and he feels it is important that they are met.


Leslie: the IPv4 outage experiment was consciously different than the experiments at NANOG and APRICOT. 

Fred: if you want to operate v6-only get on the v6 SSID. It is already happening.

Russ: it seems to be clear that a v6-only network is desired for future IETF meetings.  As the v6ops WG is struggling with the NAT-PT space..... (Russ?)

Leslie: the ipv6-nat network experiment is separate and will bring different results.

Lucy: the networks at NANOG and APRICOT included an IPv6-only network and a v4 and v6 only network...(asked Lucy what she said. She will send text)

Spencer: suggests to publish the Successful Protocols Document as an RFC

Dave Thaler: yes, that is the plan.

Sam Weiler: very disappointed that the IAB put so much pressure on an independent nomcom that they felt pressured to get an arbiter.

Erik Rescorla: the IAB has to do due diligence, and cannot just assume the nomcom is making the right decision.

Olaf: the IAB's decision and rational is included in Lakshminath's report. Suggests people look at that.

Rob Austein: having served for 6 years on the IAB, it can be quite difficult: one is often handed a slate with very little information. One can either rubber stamp or ask for more information. The rules are very fuzzy at the moment. Don't want to blame any side. Just has to be fixed.

Danny: suggests to make changes to the form to de-couple confidential information from other information.

Loa: when the IAB gets a slate it assumes the nomcom was doing a good job.  We have a situation where we have to take a decision and we need to make sure that this is a decision that we can stand behind.

Sam: looking forward to the discussion on the confirming body's role.

Leslie: suggests to review the previous nomcom WG mailing list archives

Gregory: appreciates the fact that his privacy was held so high by the nomcom, appreciates that the IAB is trying to make the right decision and also appreciates that all parties tried to do the right thing.  This is all healthy behaviour. Proud to be of a community that is behaving like that.