DECADE BoF/WG, Friday, 13:00-15:15

Decade BOF/WG Minutes
Meeting: IETF77, Friday, March 26, 2010
Location: Hilton Anaheim, Redondo room, 1300-1515
Chairs: Rich Woundy, Haibin Song
Minutes: Enrico Marocco

There were about 40-45 attendees.


1. Agenda bash
2. Problem statement draft-song-decade-problem-statement-01
3. P2P background and foundation
4. Requirements
 a. Original requirements, draft-gu-decade-reqs-04
 b. Additional requirements, draft-ohlman-decade-add-use-cases-reqs-00
5. Survey, draft-song-decade-survey-03
6. P2P resource control survey, draft-chen-decade-resource-control-survey-00
7. Conclusion and next steps (skipped later)
8. Solution architecture outline

Alexey: we are not going through the charter, as it has been approved by the IESG and will go out for IETF-wide review very soon.

There were no questions nor comments.

Problem statement

Haibin presenting.

Hannes: liability issues may be a show stopper. Haibin: it is a business issue. Barry: it has been moved out of scope for that reason. Hannes: I would check with ISPs about this.

Rolf: during the plenary they showed that P2P traffic is declining, there may be bigger fishes to fish. Mat: P2P is not shrinking, it is other traffic that is growing. Rich: The issue with P2P is uplink bandwidth utilization. Richard: There are companies putting P2P technology into their CDN solutions. Rolf: Joost started as a P2P application, now it's client/server. Richard: The traffic for P2P file sharing may be shrinking, but the traffic for P2P streaming is not shrinking.

Yunfei: How does DECADE relate to transparent caching. Haibin: applications don't have control.

Aijun: this approach requires a change in P2P protocols. Haibin: I agree it needs some change in the P2P client to use DECADE. Some P2P applications such like VoIP maybe do not have the motivation to use DECADE, but for file sharing and streaming applications, they have the motivation to use DECADE for content distribution.  

Richard: we are not necessarily talking about unlimited cloud storage. You may need a subscription to use it. Aijun: file-sharing users may not want to pay for that. Richard: it may be included with your subscription. Peter: last time I checked, the IETF did not have much experience in business models, suggest to move on.


Richard presenting.

Yunfei: what's the impact on existing systems? Users may have limited uplink. Richard: indeed, that's an issue a network storage can help address.


Yingiie presenting.

Yunfei: deleting content makes it more a storage than a cache. Yingiie: yes.

Yunfei: is it really necessary to be able to access data before it is fully stored? Haibin: the WG will figure it out.

Yunfei: access control will increase latency. Yingiie: we need to discuss this as well.

Rich: we'll skip the break.

Peter: don't see anything about identification and naming of data. Those things always imply internationalization issues.

Additional requirements

Ove presenting.

David: agrees with the idea, something similar happend in P2PSIP. A totally general solution is too complex, one with one use case only is silly. Reasonable to try to be general, but optimize for the main use case.

David: are there any implications for the protocol to make the solution general? Ove: maybe.

David: in the mobility use case, don't need to access the locker too often. Ove: right, mobility is not a strong requirement.

Rolf: worried about adding too many reqs, e.g. mobility.

Bhumip: what's your definion of IPTV? May imply copyright issues. Rich: there are issues also, e.g., for advertizing. That's why content providers not often like caching.

Yu (??): is this a full CDN solution, or just network storage? Ove: there are hybrid solutions, e.g. Spotify.


Richard presenting.

Peter: there is a WG doing OAUTH, really suggest to go talk to them as access control reqs here are very similar to theirs.

Steven: there is a lot of literature about content centric networking. Take a look at the work done in the DTN RG.

Roni: there is a lot of stuff don't see how can be useful.

Survey of resource control

Haibin presenting.

There were no questions nor comments.

Solution architecture outline

Richard presenting.

Ingemar: do not limit to P2P only. David: focus on your main use case, but try not to break other applications.

David: how can we contribute? Haibin: reqs, architecture, extensions to existing protocols. Rich: we need editing cycles. What's most important now are reqs and an accurate survey. Who can help in general? 6-7 shows of hand.