IETF 81 Quebec, 23.-29. 07.2011 
Mo 24.07.2011, 13:00-15:00 PPSP 
Minutes taken by Christian Schmidt 
Chairs: Martin Stiemerling, Yunfei Zhang
Problem Statement:  draft-ietf-ppsp-problem-statement (Yunfei Zhang) 
Martin Stiemerling: WGLC will be started next week, comments within 2 weeks expected. 
 

 >>Protocol Requirements:   draft-ietf-ppsp-reqs (Ning Zong) 
Johan: Do you mandate bitmap approach or can you making this more general. 
Ning Zong: Agreed, bitmap approach will not be mandatory. 
Martin Stiemerling: Also will go to WGLC after problem statement WGLC will be finished, meaning 3 
weeks from now on.

>>Survey:  draft-ietf-ppsp-survey (Gu Yingjie) 
Martin Stiemerling : Since we have allready 2 WGLC on the pipe, then we will start WGLC on this 
document, no new version for this needed.

>>Tracker Protocol: draft-gu-ppsp-tracker-protocol (Gu Yingjie) 
Arno Bakker: Implicit Join is not included in the ID, it is important to improve efficiency. 
Gu Yingjie : Yes, it will be included. 
Xiao Lin: Stat_Report ? meaning peer status? 
Gu Yingjie : Yes. Separate list or integrated in swarm info has not be decided yet. 
Johan Pouwelse: Timeout event is without mentioning a figure e.g. 1 minute or 1 hour.
Yingjie: Real system should be able to give us recommendations, if not we will leave it to 
implementation. 
Jana Balsa: Especially critical for keep-alive timer. 
Arno Bakker: Can you extend the draft with a threat model? Use a more detail model on what you 
expect. 
Gu Yingjie : Yes, this can be added in the next model. 
Yunfei Zhang: Connect and disconnect contains certification, there are no certifications in find and 
join ? don?t we need this as well? 
Gu Yingjie : Certificate in Find and join are optional. 
Wes:  I am not sure, if you use valid XML syntax in the ID. The messages are not XML. 
Gu Yingjie : will check. 
Martin Stiemerling: You should work on the XML adaptation. It is not very easy, if you are not used to 
do this. 
Martin Stiemerling: We have a second ID for the tracker protocol, but they are not coming to the IETF 
meetings. This ID is focused on layered streaming.  
Gu Yingjie : We will have to look, if a merger is possible. 
Martin: That?s good.

>>Peer Protocol-1:  draft-gu-ppsp-peer-protocol (Gu Yingjie) 
Arno Bakker: ID should be extended to make this secure. Currently there is no protection to get back 
IP address e.g. from the White House. 
Gu Yingjie: Accepted, will be added. 
Arno Bakker: In HTTP the described kind of Push is not possible; this is not the best choice here. 
Arno Bakker: It is also important to look at latency here for selection of proper transport protocol. 
End user latency should be important issue for this decision. 
Gu Yingjie: Proprietary messages could be added later on by the implementers. 
Johan Pouwelse: Strongly recommend a complete specification, not leaving too much space for 
implementation choice. And try to avoid redundant messages. 
Wesley: Why do you want to use HTTP, not just opening a channel? 
Gu Yingjie: I want to make it consistent with the tracker protocol. 
Wesley: For the tracker protocol the same question, why do you want to user HTTP? 
Gu Yingjie: It was a WG decision. Can you give suggestion for alternatives? 
Wesley: Most of the encodings for HTTP processing are not there. First figure out the messages, you 
need, then select a proper transport protocol. 
Johan Pouwelse: Pleading for UDP without HTTP and binary encoding.
Martin: Chair clarification: Peer protocols have not yet made progress as expected. We are trying to 
make more push there. HTTP should not be the solution we should use for this purpose.  In case of 
UDP we could have some kind of congestion control.

>>Peer Protocol-2: Swift: draft-grishchenko-ppsp-swift (Arno Bakker) 
Zong Ning: Charter do not include data transmission. You can explain why this is done here. Anyway, 
you can use HTTP for this purpose as well. Chairs please clarify what?s in/out scope. 
Yunfei Zhang: PPSP only focus on signalling according to the current charter, transmission part is not 
included. If there is strong consensus in the WG, we should discuss charter update. 
Zong Ning:  PPSP is more about signalling, you are here talking about data transport, this is not in the 
current scope of PPSP. 
Gu Yingjie: Have you defined the bitmap for live streaming? It is not good for file sharing. 
Arno Bakker: For File sharing we would need another bitmap for file-sharing. 
Yunfei Zhang: Is swift transport independent? 
Arno Bakker This has been included in the protocol. 
Yunfei Zhang: I really like this proposal, for me it is not important, where this will be done.

>>NAT traversal:  draft-li-ppsp-nat-traversal (Lichun Li) 
Johan Pouwelse: According to our node measurement experience, this could be simplified. Take our 
research results as input for your work. You can do everything with a single primitive.

>>eICP:  draft-zhang-ppsp-eicp (Christian Schmidt) ? only 3 minutes left. 
Main goal for enhanced ICP: Usage of network caches for PPSP. 
Main issues to be solved: Caching of p2p traffic and chunk addressing. 
Martin: Should be clarified, how this relates to the work in DECADE.