Nice shirt
", "time": "2024-11-05T16:50:34Z"}, {"author": "Tobias Fiebig", "text": "All the pufferfish in line?
", "time": "2024-11-05T16:56:25Z"}, {"author": "George Michaelson", "text": "when I measured, RPKI signed objects were too small in the main to leverage Filesystem block by block checksum tests in rsync to do partial file send
", "time": "2024-11-05T17:01:05Z"}, {"author": "George Michaelson", "text": "which is how rsync can be more efficient. it requires files which span disk blocks, to make block checks worth doing
", "time": "2024-11-05T17:01:46Z"}, {"author": "George Michaelson", "text": "\"put it in BGP\"
", "time": "2024-11-05T17:02:02Z"}, {"author": "Jeffrey Haas", "text": "hiss
", "time": "2024-11-05T17:02:11Z"}, {"author": "George Michaelson", "text": "(this is surely what extended messages in BGP would be good for, doing flood fill)
", "time": "2024-11-05T17:02:17Z"}, {"author": "Jeffrey Haas", "text": "w.r.t. TLS as transport, I'd think that \"doesn't rely on webpki\" is probably something that belongs as a sub bullet.
", "time": "2024-11-05T17:05:46Z"}, {"author": "George Michaelson", "text": "its only 80,000 clients. NOTIFY outbound should be possible. its not like it has 1b edge users, 80,000 is not going to flood protocol control block space to be connected
", "time": "2024-11-05T17:07:29Z"}, {"author": "George Michaelson", "text": "(ok.. fermi maths 1 client per ASN)
", "time": "2024-11-05T17:07:40Z"}, {"author": "George Michaelson", "text": "Thats George and Byron Rudiger.
", "time": "2024-11-05T17:09:20Z"}, {"author": "George Michaelson", "text": "it was basic gzip expansion bombing type situations in the main
", "time": "2024-11-05T17:09:41Z"}, {"author": "Jeffrey Haas", "text": "the gzip thing is a good example of why transport integrity is a necessary property.
", "time": "2024-11-05T17:10:32Z"}, {"author": "George Michaelson", "text": "I think there was also a go-slow, and an unavoidable problem with a publication URL which is a machine which emits 1m ROA and does it slowly, and everyone is going to go there and like I said its only 80,000 PCBs so its not like you need a giant farm to slow the fetcher down.
", "time": "2024-11-05T17:11:30Z"}, {"author": "George Michaelson", "text": "(hand up for when he's done not seeking an interrupt)
", "time": "2024-11-05T17:25:09Z"}, {"author": "George Michaelson", "text": "(and if there's no time I wont complain.. much)
", "time": "2024-11-05T17:25:46Z"}, {"author": "George Michaelson", "text": "I think Job may overstate \"can never happen\" because companies can relocate resource management without transfers. So there is a context where 44/8 can appear in APNIC.
", "time": "2024-11-05T17:28:23Z"}, {"author": "George Michaelson", "text": "he's right transfers cannot cause it, but transfers are not the only reason resource management moves
", "time": "2024-11-05T17:28:44Z"}, {"author": "Chris Morrow", "text": "How does this contaraints plan work with transfers across RIRs?
", "time": "2024-11-05T17:28:52Z"}, {"author": "Randy Bush", "text": "amazing rigitity and complexity to cover the crap caused by the rirs blocking the iana issuing certs to the rirs
", "time": "2024-11-05T17:29:40Z"}, {"author": "Bob Beck", "text": "Doesn't that then have the problem that the RFC 3779 objects need to drool all the way up the intermediates?
", "time": "2024-11-05T17:30:38Z"}, {"author": "George Michaelson", "text": "@bob issuer subject name paths in path building to a TA does have to happen if you validate from the edge, but in fetch logic, top down, its simple, most rpki clients work top down. not saying its problem free. just observing mostly it doesnt arise. but alternate paths issuer-subject are a fact of life in PKI
", "time": "2024-11-05T17:31:44Z"}, {"author": "Bob Beck", "text": "your openbsd mechanism avoids the intermediate problem by only applying to end-entitiies
", "time": "2024-11-05T17:31:47Z"}]