IETF-76 P2PSIP Meeting notes (DRAFT)
Note takers Jim McEachern and Spencer Dawkins. Edited/compared with audio by David Bryan.
REsource LOcation And Discovery (RELOAD) Base Protocol,
Cullen Jennings,
Most points for discussion on slides are cosmetic/minor editorial, with a few exceptions.
The draft is completely wrong for calculating the signatures. Proposal to use Michael ChenŐs suggested change. Group indicated that this path was supported.
Turn density. The system needs an algorithm for this to work, and this approach works, even though it is almost trivial. Propose to just leave it as is since the system is not very sensitive to this (see below). However, he will update the draft to document some of the limitations of this approach and to point to better algorithms that might be considered in future work. David Bryan asked from Jabber room about the slide stating it should be used since studies and experiments indicated it worked well enough and was robust if you got it wrong.
David noted this assertion had been controversial before, and asked where he could find these studies indicating it worked, and Cullen indicated that the authors had not shared them publicly and felt it was too much work to publish the results. Indicated that there was an opportunity for more work in this area for a general service discovery algorithm.
Self Tuning: Proposal to include information on successors and predecessors in Leave messages as it is very helpful for the work in the self tuning draft. Cullen was interested in what the group thought. Comment in favor of including this as a should. Cullen will do that for the next draft.
Other Issues. None
Robert raised Ňopen issuesÓ in the document.
Reactive recovery. People keep suggesting they will provide input, but they donŐt provide anything. Cullen is therefore proposing to delete this as an open issue.
CullenŐs view is that once he updates this document with these changes and there is time to review, this will be ready to go to a WG last call. Consensus of the room was to do so.
David Bryan asked substitute chairs to get a list of reviewers who would do a full review of the document:
Looking for detailed reviewers
P2PSIP Security Overview and Risk Analysis
Song Haibin
draft-matuszewski-p2psip-security-overview-01
Presentation outlined the two changes that were made to the presentation. Asked if there were any additional comments.
Cullen said that he had trouble commenting because he was unclear as to exactly what the purpose of this document was. Without that, it is hard to comment.
From Jabber room David mentioned that in an earlier meeting the consensus was for this to provide guidance to people new to P2P about the unique security issues and implications.
An extension to RELOAD to support Direct Response and Relay
Peer routing
Ning Zong
Comparison of DRR/RPR vs. SRR and the number of messages and hops. Questions from Cullen about exactly what is being analyzed since the number of messages seems low to him, especially if they include the entire TLS handshake. Roni Even says that it does include the TLS handshake. Cullen is unconvinced. Agreed to take this offline to investigate further.
The authors feel they have addressed the comments and that
it is an optional method for particular deployment scenarios.
Load balancing models for DHT-based Peer-to-Peer Networks
Erikki Harjula
draft-harjula-p2psip-loadbalancing-survey-00
Load balancing is critical, but DHT does not achieve acceptable load balancing. Therefore more analysis is needed.
Most techniques use:
- measure load
- distribute load information
- balance the load
Of the many methods, they focus on four.
Summarized a brief analysis of the attributes of each
method, including cost in that analysis.
This analysis is very tentative, and they plan to extend the analysis to
provide significantly more detail.
Only two people have read the draft
A Self-tuning Distributed Hash Table (DHT) for REsource
LOcation And Discovery (RELOAD),
Jouni Menp,
draft-maenpaa-p2psip-self-tuning-01,
Previous version did self tuning and load balancing. Current version is self tuning only.
With static parameters approach it is not possible to have both low stabilization overhead and low failure rate.
Self tuning allows parameters to change. Each peer collects data and uses this to dynamically adjust parameters.
Question to the group as to whether or not the group would be interested in having a milestone related to self tuning. Support was expressed for this work, but not that many people have read it. Jon encouraged the work to continue, but with such limited audience, was reluctant to adopt it as a work group. Jouni countered that we had that situation at the last meeting and that if it became a WG item, then perhaps more people would actually read it. Extended discussion, with everyone generally supporting this work.
Jon asked how many people understood the problem that this addressing. About 20-30 people raised their hands.
Poll: How many people think that the WG should have a charter item to address this problem? Result - Audible support and no objections.
Poll: Should this draft be used as input into that charter item? Result - Audible support and no objections.
Jon said they would pass this along to the ADs for consideration.
Service Discovery Usage for REsource LOcation And Discovery
(RELOAD)
Jouni Menp,
draft-maenpaa-p2psip-service-discovery-00
Outlines a proposal for a generic service discovery mechanism
Poll: Should we be defining a generic service discovery mechanism for p2psip? Result – lukewarm interest, with no objections.
Conclusion. Encouraged to continue working on this and bring it to the list to continue generating interest.