sidrops 112 Monday 8 November 2021 12:00 UTC Agenda 1. Chair slides - Chris Morrow Alexander Azimov: both ASPA drafts still need some work before we can request WG last call 2. Ben Maddison: https://github.com/benmaddison/rpkimancer Job Snijders: Expresses his appreciation Chris Morrow: Being able to look at the differences between objects over time is useful. Ben Maddison: Being able to do a diff between two known versions of a given object is not difficult to implement, but itÕs difficult to scale storing that data. This tool is not a good place to have that. Chris Morrow: Perhaps a thought process on the mailinng list on how to store that historical date. Ben Maddison: The issue is that you end up with a massive duplication from rollover manifests and crl , tracking at file level would make a diff Ôalmost freeÕ but a bigger job than a small python library like this (comments mentionÊhttps://rpkiviews.org, RIPE NCC RIS data) Job Snijders: In order to track what happens Òin the middleÓ we need Certificate Transparency. Thats perhaps a topic for the next IETF meeting. Chris Morrow: slideware/discussion welcome Ties de Kock (chat): My experience is that if you want to actually test RPs having a valid tal + repo (so including valid URLs) is likely the way to go. Needs one implementation instead of N. Otherwise this looks like a useful tool! The RIPE NCC also has data per TAL (both effective VRPs and the raw objects) since 2011. Ben: I agree in genral, hooking up a test RP to this data is probably easiest to implement if you use one of the retrieval mechanisms like RRDP or rsync running locally, but the problem is that because you need to embed the URLs in the actual objects themselves, you need to spoof your own machines DNS, feels a bit clunky. Doing both is probably the right answer. 3. Oliver Borchert: BGP-ASPA Hackathon Report Tools, Datasets to facilitate testing route leak mitigation https://github.com/usnistgov/NIST-BGP-SRx Ruediger Volk: Syntax qestion: the syntax that you give for the ASPA test data ends in an asterisk sign, should it be a plus? Oliver Yes. Ruediger: That precludes empty AS provider sets but I hope they are still in ASPA. Oliver Borchert: ok weÕll change back to plus so its regex zero-or-more. Alexander Asimov: I have a question about the source of data. Using CAIDA, As I remember, peering relations based on ÒexpertÓ set of tier-1 providers. These T1s are not in the customer-to-provider set. you need to specifically create the ASPA0. IÕm surprised by number of unknownsÉ Oliver Borchert: Take the data with a grain of salt. DidnÕt add time analysing the data. For the scale test, and some dataset which would make some kind of set instead of random. Want data which can be used to do Òtwo implementations, output the sameÓ to validate the algorithm is deterministic. Alexander Asimov: You should check how you are processing data from CAIDA. It is known to be noisy and have false positives and there are missing providers. I suggest to start with simple ASPA records forTier 1 providers. Oliver Borchert: We needed a big dataset, weÕre open to other datasets. Sriram Kotikalapudi : Small data testing is available on the github page. Alexander Asimov: I observed that the number of unknowns is large, we noticed that they should have an AS0 ASPA. The 701 AS is a Tier1 appears as a customer, that is unusual. From the draft there is no Òempty ASPAÓ (Alexander yes in the draft this is marked by AS0 but discussion on the list continues) Oliver Borchert: will check the PDU, if mistaken will modify, deal with this Doug Montgomery: (from the chat) Yep - our goal was just a large data set and a common APSPA output format - run two ASPA implementations on the data set, diff the output - algorithm ÒinteroperabilityÓ. It is not meant to be a realistic test of ASPA effectiveness. Chris Morrow closes the meeting at 13:21 UTC