IETF 117 SAVNET Meeting minute

1.	Dan Li “A summary of intra-domain and inter-domain SAV problem statements and next-step work”

Joel: clarify the status of bulletin points: the first two and the second two points are different.
Jeffrey Haas: Absent from requirement: SAV information propagates different from routing information. Out of sync between source validation and forwarding change. How to minimize this? 
Joel: Convergence discussed but not clear. Discuss in the list.  
Dan: Talked about this last time. Will have a presentation to show the analysis work.  Important but no sure if it should be included in the problem statement draft.
Rudiger Volk: what you consider accuracy?
Joel: Draft talks about the inaccuracy and the problems it caused.
Rudiger: Related to the convergence question. People think accuracy is binary. Avoid damaging the gray area. False positive and false negative are insufficient to chart the space. 
Joel: Please point out how to improve the document.
Ben Madison: Convergence is hard. More important, semantics of the information relate to the forwarding database. Race condition may introduce black hole.
Joel: Make sure we agree on the protocol behavior regarding to the convergence and race condition.
Ben: No tight coupling. Live with the transient mismatch and don’t do harmful things.
Dan: will be covered in the protocol design.

2.	Nan Geng “SAVNET architecture for intra-domain network”

Jeff Haas: addressed to chairs. Architecture contains what I asked for. But mismatch with the requirements. What to do about it? 
Joel: If you find any, tell us.

3.	Lancheng Qin “SAVNET architecture for inter-domain network”

Joel: make sure the architecture proposal is clear.
Rudiger: whether or when a basic model of SAV specific message content is introduced?
Joel: will need that.
Lancheng: will do but not in the architecture draft.
Ben: SAV specific protocol should not be used to distribute information for forwarding.
Igor: The draft seems to bifurcate information into sav specific. Think about the possibility for dual usage protocol. Static and dynamic information 
are treated the same. Easy to sync static information for inter domain. Probably make distinction in the doc.
Jeff: should discuss what incremental means in deployment considerations. 
Lancheng: will consider it.
Joel:  need to distinguish between incremental deployment of support for the SAV protocol and its information incremental deployment of acting on the SAV information.
Sriram: You will include the forwarding path in the SAV specific message. Does it contains an AS path?
Joel: Need to see the information model. Can’t answer it right now. Need to clarify the draft.
Lancheng: will submit another draft on the content and data structure.
Joel: maybe just information model. We need to know what it contains. 
Linbin: talk about incremental deployment issues.

4.	Lancheng Qin “How much larger is SAV table compared to FIB? A study with real BGP data”

Jeff Haas: any prefix aggregation or all prefixes?
Lanchange: data from routeview without modification 
Jeff: like to see the aggregation effect in the next version.
Ben: The implication is one can inspect routing data and determine feasible rpf path. If so, just use BGP. Don’t think that the study can be done with this data.
Rudiger: Assumption that just one selected path is used is not true. 
Lancheng: we made some assumptions here
Nan (chat): A comment on the SAV table size analysis: From Some perspective, the results are also overestimated because allowlist should only be used at customer/peer interfaces instead of provider interfaces. But in the analysis, the latter case seems to be considered, which will result in large table size.

5.	Libin Liu “Recent Progress of SAV Open Playground”

R: this is really neat. Want to play with it.

6.	Fang Gao “A Yang Model for SAVNET”

Joel: No comments at this time. It's very early to figure out the YANG structure. 
Ben Schwartz: mailing list for a radical different approach on dataplane. Talk to me if interested.