P2PSIP Meeting, Thu Mar 25, 13:00-15:00 ======================================= (Minutes by Enrico Marocco with edits by David Bryan) Agenda bashing -------------- Note Well Presented. No questions nor comments. RELOAD Open Issues ------ Bruce Lowekamp presenting. Slide 8 (Overlay Link Protocols) David Bryan: When negotiating with the different mechanisms like DTLS TLS-ICE or whatever, is there some error handling for bad cases? We need that. Cullen Jennings: didn't we say that we can't have ice and non-ice in the same overlay? Bruce: it should work. Cullen: OK, if I have a problem I'll send it to the list. Bruce: how many people have read the last stuff about transport? One show of hand. Roni Even: I started to read the draft, but found many inconsistencies and decided to wait for a new version. Bruce: we've fixed many, if there are any left, we'll fix them. Slide 12 (Direct Return Response) Roni: direct return response is one of the inconsistencies, there is an undefined parameter. Appendix B is contradicting. Bruce: still not clear whether DRR should be in the base draft, or in an extension. Brian Rosen: the authors propose that the base draft could provide a socket for the extension draft to plug in. Does this work for you? Roni: it's highly unclear. Bruce: base draft can describe the mechanism, the extension draft could provide guidance and rationale. Brian: suggest that Roni works with the authors. Cullen: connection information is defined, even if not in the clearest way. Roni: ah, OK. Brian: bring the discussion to the list. Slide 13 (Overlay Algorithm) Henning Schulzrinne: how common is it not to ask for finger table entries? Should always send them at Attach time. Bruce: something like adding a code point in Attach? Henning: yes, but don't really care about how, as long as you do that. Jouni Maenpaa: I support Henning's point, as it help reduce the overhead when churn is high. Hum on three options for doing attach options Choices: 1. leave as is; 2. add a flag; 3. have a flag for future use (Henning's proposal) Result: Weak consensus for 2 (add flag to attach to request update on resp change) Slide 14 (TLS Mandatory) Brian: if don't want TLS mandatory, speak up now. Roni: is it mandatory to implement, or mandatory to use? Bruce: that would be an overlay configuration. Eric Rescorla: just to be clear, if you don't use TLS, you must some other security mechanism. David: concerned about new security mechanisms that would break. Kostantinos Birkos: I'm working on different security, if you don't use TLS, is still encryption mandatory? Bruce: can't tell, should go through this in detail. Slide 15 (Peer-ID Length) Henning: Should not limit the length, be future proof. What's the benefit of enforcing a maximum? Eric: it's practical to have fixed size IDs. Cullen: can live with variable size, but want a max. The ID is an hash, should not be anything else. There is nothing out there that uses more than 160 bits. David: would like to have fixed size per overlay. Hum on three options for Peer-ID Lengths. Choices: 1. 128; 2. per overlay variable, max 160; 3 per overlay variable: Result: Slight consensus for 2, will take to the list. Henning: what would happen if someone has a magic DHT that requires more than 160? Brian: will take to the list. Conclusion: Bruce: will iterate soon and will provide a really really final version, ready for WGLC. David: please look at this when it comes out. Self-tuning DHT --------------- Jouni presenting. David: do you expect to have to do changes when the base document changes? Jouni: Little. Cullen: is there anything we should change in the base draft? Jouni: no. Brian: is this draft adding a keepalive mechanism? Is that useful for other things? If not is fine where it is, otherwise it should go in the base draft. Diagnostics ----------- Haibin Song presenting. Cullen: what extensibility do you need in Ping? Haibin: there is no extensible parameter. Cullen: all messages are extensible, it's the message extensibility mechanism. Eric: section 5.3.3 defines the extension mechanism you are probably looking for. Otherwise explain why it doesn't serve the purpose. David: This changed during a recent revision. Need to revisit and see if changes meet needs. Roni: in previous meeting there was consensus for keeping Ping simple. Don't have an opinion whether should be extended Ping, or a new Ping. David: we still want to know whether the group wants Ping or something else. In Minneapolis it seemed a controversial issue. Brian: will ask on the list, not the right time to ask it now. Haibin: should we use Route_query plus Fetch (in the base draft), or Path_track (new)? Cullen: didn't we decide to define a mechanism to make Route_query and Fetch atomic? David: after Hiroshima people claimed that making Route_query and Fetch atomic is an overkill. Hum on two options for path_track like behavior. Choices:1. Route_query+Fetch; 2. Path_track. Result: Few options, will go on the list. Brian: If we don't hear much we'll look toward Path_track. Cullen: hope people review the document thinking of what information could be added and be useful. Security mechanisms and key refresh ----------------------------------- Kostantinos presenting. Brian: during key refresh, if the previous key was compromised, the new key will be compromised as well. Cullen: we don't have any security people. I understand the need for key refresh. CMS would be probably the way to go for encryption. That the sort of considerations that would like to see in the draft. Kostantinos: focus was on the topology information. Bruce: how does the info protection work? If you simply observe messages, you can pretty much get the structure. Cullen: interested in location privacy. There is probably a chance to build a TOR-like system, but need rechartering. Brian: keep the discussion going, iterate, reflect comments. Wrap-up ------- Brian: we are 80% done, do not let the attention drop. Let's this finish this stuff, we can do WGLC of the base draft by the next meeting. David: until we are not done with the base draft, there will be no room for new work. If you have other interests, help finish the base draft first.