Show Idle (>14 d.) Chans


← 2022-01-23 | 2022-01-25 →
03:23 * asciilifeform massaged 3.2.2 and 4.3.1, and elsewhere; stops for the day
~ 1 hours 32 minutes ~
04:55 asciilifeform thimbronion, jonsykkel , et al : had possibly interesting idea for auto-resolution of forks. record the first l1 peer via whom you received a hearsay bearing that speaker. the chain of the fork received via that peer is to be considered genuine.
04:56 asciilifeform applies (unless asciilifeform miscalculates) recursively, to enforce '1st user owns the handle' across the pestnet.
04:57 * asciilifeform invites folx to come up w/ a counterexample, in which algo bombs
05:00 asciilifeform http://logs.nosuchlabs.com/log/asciilifeform/2022-01-23#1075817 << in the absence of an immediate message or prod from that speaker (if he is in-wot) , that is.
05:00 dulapbot Logged on 2022-01-23 23:55:51 asciilifeform: thimbronion, jonsykkel , et al : had possibly interesting idea for auto-resolution of forks. record the first l1 peer via whom you received a hearsay bearing that speaker. the chain of the fork received via that peer is to be considered genuine.
05:02 asciilifeform ( a message directly from a peer already resolves any forks afflicting that peer)
05:02 dulapbot Logged on 2022-01-21 18:00:55 asciilifeform: thimbronion: a msg from a direct peer is prima facie authentic
05:07 asciilifeform on contemplation, not clear that forks can even occur at your station if all of your peers apply this rule.
05:08 asciilifeform could dispense, then, with the fugly 'fork alarm' thing entirely.
05:08 asciilifeform (and if one of yer peers ~isn't~ applying it, will be immediately clear ~who~)
05:09 asciilifeform this might be the sought-after finalsolution, unless asciilifeform is missing sumthing.
05:09 * asciilifeform to bed.
~ 33 minutes ~
05:43 jonsykkel im not clear on wat a fork is exactly, prev understanding was that fork = 2+ msgs exists with same selfchain value (chain has fork shape to it), and can occur for peers as well as non peers
05:43 jonsykkel reading draft now, understanding is that fork has occured if recv msg that overlaps with previously recorded chain in time dimension (or msg that predates genesis, presumably), and is strictly relevant for non peers
~ 1 hours 43 minutes ~
07:27 jonsykkel http://logs.nosuchlabs.com/log/asciilifeform/2022-01-23#1075817 << say u have following net: http://zzz.st/up/vFniopMQ/20220124_081722.png where square=pestron, color=handle, line=peering. noone has ever sent any msg, then blue(a) broadcasts a msg. yellow guy records msg and relays to red and green. red records msg. green rejects msg cuz in-wot hearsay (draft 4.3.1 step8). now blue(b) broadcasts
07:27 dulapbot Logged on 2022-01-23 23:55:51 asciilifeform: thimbronion, jonsykkel , et al : had possibly interesting idea for auto-resolution of forks. record the first l1 peer via whom you received a hearsay bearing that speaker. the chain of the fork received via that peer is to be considered genuine.
07:27 jonsykkel msg, green records and relays to red and yellow. red autoresolves cuz considers blue(a) the genuine blue chain. yellow rejects cuz in-wot hearsay. is this correct understanding? wat happens next, how does anyone know theres a collision?
~ 6 hours 38 minutes ~
14:05 asciilifeform jonsykkel: asciilifeform reverted the 4.3.1 step8 thing before went to bed
14:07 asciilifeform green & yellow will not reject apriori
14:12 asciilifeform they'll have seen blue(a).
~ 1 hours 27 minutes ~
15:40 asciilifeform jonsykkel: 1 possible issue (in light of the current buffer logic) is if blue(a) and blue(b) fire within 'Te' interval.
15:42 asciilifeform if they do not, and, say, blue(a) fires, all three (yellow, red, green) will store 'yellow' as designated relayer for 'troo' 'a', and worx
15:42 asciilifeform ditto if blue(b) fires 1st, then store 'green'
~ 21 minutes ~
16:04 shinohai $ticker btc usd
16:04 busybot Current BTC price in USD: $34105.7
16:04 shinohai $uptime
16:04 busybot The bot has been up for: 52 days 29 minutes and 40 seconds
~ 19 minutes ~
16:23 asciilifeform 34$ when!
16:25 thimbronion asciilifeform: let's say the peer via which first hearsay message from alice goes offline. Now we receive a message from alice that breaks the chain. Alice is permaforked?
16:27 asciilifeform forked until $peer returns and relays the original's; or until resolved by %resolve (which loox like you'd still need, ugh)
16:29 * asciilifeform quite convinced that there's a clean solution to all of this, simply asciilifeform somehow missing it, not looked in correct direction evidently
16:30 PeterL would you only be able to "getdata" from peers, or can you ask for intervening messages from L2?
16:30 asciilifeform only peers
16:30 asciilifeform getdata aint relayed.
16:31 asciilifeform (it's potentially expensive)
16:31 PeterL although, if you are seeing messages from L2 then they must be connected to one of your peers, presumably they would have the intervening messages?
16:32 asciilifeform not necessarily
16:32 asciilifeform for instance, they would not have seen messages preceding their peering
16:32 asciilifeform (unless successfully getdata'd upstream for'em)
16:32 PeterL so not guranteed
16:33 asciilifeform aha. in general getdata aint guaranteed to succeed.
16:33 asciilifeform (tho often enuff you'll have entire msg history of the pestnet from genesis. but not guaranteed.)
16:38 asciilifeform orthogonally, asciilifeform still considering variant of this, but corrected -- the idea is, one ought not to process an in-wot hearsay if the speaker corresponds to a 'warm' peer. (you'll either defo get the immed. copy, eventually, or oughta prod him then getdata for it until you do)
16:38 dulapbot Logged on 2022-01-24 09:06:12 asciilifeform: jonsykkel: asciilifeform reverted the 4.3.1 step8 thing before went to bed
16:39 asciilifeform given addr cast, there oughta be no such thing as a 'broken peering' while peer's station is standing
16:39 asciilifeform (if it aint standing, anyffin you see w/ that speaker is ipso facto bogus)
16:40 asciilifeform i.e. not clear that you'd want to process in-wot hearsay at all.
16:41 PeterL the other messy time is if A and B each have a different peer named C, how do things get handled if A and B peer?
16:43 asciilifeform well, ideal would be if somehow distinguishable.
16:46 asciilifeform w/ the current logic, doubt that A and B would stay peered for long, unless one of'em is willing to give his C the boot.
16:49 asciilifeform let's turn the problem 'inside out' and give ~desired ultimate condition~ (in the same way the 'martian bank' describes the desired condition of bitcoin net) .
16:49 dulapbot (trilema) 2016-10-20 asciilifeform: ('martian bank' being simply a naive abstraction of 'idealizes swiss bank', where money supply is constant, and i can send from account a1 can send to a2 if and only if i have the privkey for a1, and double-spend - impossible, etc.)
16:51 asciilifeform what asciilifeform wants is to emulate the 1 desirable aspect of a traditional centralized 'fleanode' : that the 1st user of a handle on a given pestnet can be distinguished from subsequently appearing users.
16:51 asciilifeform this is likely impossible in the general case.
16:52 asciilifeform q is -- to what extent can one in fact square this circle.
16:56 asciilifeform ... suppose, for instance, that the msg having ~lower min bounce~ is considered (in absence of immediate msgs from $speaker) to represent the troo branch of a fork.
16:56 thimbronion asciilifeform: could the original handle user broadcast a pubkey which could be used to answer challenges if an imposter shows up later? Or is this what we already have with resolve, sorta?
16:56 asciilifeform thimbronion: no pubkeys in pest.
16:56 asciilifeform you can of course talk to people 'out of band' and send gpggrams, etc
16:57 asciilifeform but this doesn't help a pest station distinguish fork ends for marking.
16:57 asciilifeform gotta have a means for actually suffixing e.g. bozo-1, bozo-2, when receiving l2+ hearsays.
16:58 asciilifeform thimbronion: recall the orig. reason why pest is using symmetric crypto, btw
17:00 asciilifeform ( we aint got the req'd holy grail. )
17:00 dulapbot Logged on 2021-07-03 13:14:52 asciilifeform: gold standard would obv. be rsa signatures. but these aint practical at line-rate on pc sadly.
17:02 asciilifeform ~anyffin~ that can be thrown at a station gotta be processable (incl. ~all~ K(S) attempts over whole, wot, ~and~ dedup.) in the avail. interval per packet to maintain line rate.
17:02 dulapbot Logged on 2021-09-23 15:21:07 asciilifeform: i expect you'd have at least 8cores on reasonable box, which gives you ~105840 ticks to process a packet, imho entirely doable.
17:03 thimbronion why would challenges occur at line rate?
17:03 asciilifeform not 'would', but ~could~
17:03 asciilifeform anything that ~can~, must be considered to ~will~
17:05 thimbronion challenge sould be symmetrically encrypted and manually initiated. Response would be manual.
17:05 thimbronion *would
17:07 asciilifeform on top of which, asciilifeform ~likes~ that there aint long-term keys in pest. pubkey crypto oughta be manually operated, i.e. consciously triggered at appropriate occasions, rather than sumthing a rando can cause you to do when he pleases
17:07 dulapbot Logged on 2022-01-18 16:20:27 asciilifeform: but 'erry ln you type to a stranger in a chat' aint 1 of'em
17:08 asciilifeform and there's no place to bolt on a 'manual' challenge, cuz again, no way to 'attach' pest msgs to a longterm key w/out forcing its use in sealing ~erry~ msg.
17:10 asciilifeform the fundamental unsolved problem atm is in fact ~distinguishing~ multiple l2+ stations speaking w/ same handle.
17:13 asciilifeform thimbronion: does this make sense to you ? pgpgrams to/from randos who are 'sharing party line' are completely meaningless.
17:13 thimbronion right - no way to block the l2 peer that didn't respond to the challenge - only the l1 peer that is the conduit
17:14 asciilifeform they aint distinguishable !
17:14 thimbronion l1 peers distinguishable, l2 peers indistinguishable
17:14 asciilifeform aha
17:14 asciilifeform that's the boojum
17:17 asciilifeform atm the only things which are distinguishable are 1) msgs from l1 peers 2) arbitrary msgs w/ distinct speakers 3) ~relayers~ of msgs from l2+ peers 4) chains.
17:17 asciilifeform err, 3) ~relayers of msgs from l2+ speakers
17:18 * asciilifeform thinks he's finally cleaned up all the gaffes in spec where referred to a l2+ station as a 'peer'
17:18 * asciilifeform may've missed 1 or 2 tho
17:20 thimbronion So the only person who could identify a counterfeit station would be the original station - because, hey - that wasn't me!
17:20 asciilifeform correct, that's the only operator who immediately knows (supposing he's 'warm'. if he's 'cold', i.e. he's out mountain climbing and/or his station is down, then not him, either)
17:24 asciilifeform and (as PeterL pointed out upstack ) not all collisions feature an obvious 'genuine' and 'impostor' station from the pov of all stations on the pestnet
17:24 dulapbot Logged on 2022-01-24 11:41:33 PeterL: the other messy time is if A and B each have a different peer named C, how do things get handled if A and B peer?
17:31 asciilifeform thimbronion, PeterL , et al : see the 1st thrd re subj also.
17:31 dulapbot Logged on 2021-09-13 15:04:51 asciilifeform: punkman: the stick gotta be both long and short. hot and cold.
17:31 asciilifeform ( featuring certain of asciilifeform's prev. attempts at solutions )
17:33 asciilifeform ( e.g. this one. which imho nogood. )
17:33 dulapbot Logged on 2021-09-13 15:10:22 asciilifeform: either that, or -- from asciilifeform's chalkboard -- erry message includes 2 additional 256bit fields . one is H256(R) where R is a rng turd. the other is the R from previous.
17:33 dulapbot Logged on 2021-09-13 15:10:44 asciilifeform: a message not carrying correct R is immediately known to be fraudulent.
17:33 dulapbot Logged on 2021-09-13 15:11:37 asciilifeform: this scheme is however vulnerable to mitm, anyone who is able to read a message's successor can generate a variant message, because knows the requisite R.
17:38 asciilifeform ^ in which thread, asciilifeform came up w/ interning hearsays and counting the relaying peers, to allow ~some~ distinguishing mark.
17:38 asciilifeform it's a half-solution, that one could perhaps live with, but imho still sux.
17:39 * thimbronion needs moar modafinil
17:39 asciilifeform thimbronion: the 1 from that shop you linked last yr was a++ delish, btw.
17:40 thimbronion asciilifeform: indeed no duds
17:40 asciilifeform they had an interesting spampaypal scheme for fiatola intake btw (chances are that thimbronion not seen it). they send you to some shithole and 'buy tshirts'
17:41 * asciilifeform was surprised that it wasn't a straight blackhole
17:43 thimbronion lol no didn't see it
17:43 asciilifeform prolly not long lived
~ 22 minutes ~
18:06 * asciilifeform has been bashing head against the 'fork' thing since july, not come up w/ anyffing palatable
~ 1 hours 45 minutes ~
19:52 asciilifeform ... thinking about it, this might not be a problem given that just about erryone will see the original 1st.
19:52 dulapbot Logged on 2021-09-13 15:11:37 asciilifeform: this scheme is however vulnerable to mitm, anyone who is able to read a message's successor can generate a variant message, because knows the requisite R.
19:53 asciilifeform so, for the sake of argument: suppose that (at the expense of payload) a message gets 2 new h256 fields : call'em Q and R ;
19:54 asciilifeform Q is H(message, incl. A; and a 32-byte turd, S, not included in the msg.)
19:55 asciilifeform A -- is the S from that station's prev. message .
19:55 asciilifeform err, Q and A.
19:56 asciilifeform lol
19:56 asciilifeform let's repost:
19:56 asciilifeform so, for the sake of argument: suppose that (at the expense of payload) a message gets 2 new h256 fields : call'em Q and A ;
19:56 asciilifeform Q is H(message, incl. A; and a 32-byte turd, S, not included in the msg.)
19:56 asciilifeform A -- is the S from that station's prev. message .
19:59 whaack howdy
19:59 asciilifeform wb whaack
19:59 whaack ty
19:59 whaack !e view-height
19:59 trbexplorer block_height: 720224
19:59 trbexplorer mins_since_last_block: 9
19:59 whaack !e view-block 720224
19:59 trbexplorer http://paste.deedbot.org/?id=5piY 1 of 1
20:00 whaack !e view-tx 720224 1234
20:00 trbexplorer http://paste.deedbot.org/?id=rDfM 1 of 1
20:02 whaack random tx lotto hit a segwit :(
20:02 asciilifeform lol
20:03 * asciilifeform can't tell whether ^ represents working eater or no
20:03 asciilifeform ... to wrap up upstack: if H(M_prev + A_current) != Q_prev : you have a bogon.
20:06 asciilifeform if one could fit e.g. a lamport sig in there, there'd be no headache.
20:06 asciilifeform but can't.
20:07 asciilifeform i.e. somebody who knows A_current can forge a bogus M_prev.
20:09 asciilifeform the $64 question is whether this is actually a problem , given that the genuine M_prev reaches errybody 1st.
20:17 asciilifeform the latter aint guaranteed tho, esp. if >1 msg sent in reasonably quick succession. dr.evil takes A_current and forges M_prev, and sends it out (with a doctored timestamp, for good measure), this may reach some stations on the pestnet before the genuine m_prev.
20:20 asciilifeform so this scheme's a dead end, unless somebody can think of a mechanism whereby knowledge of A_current aint sufficient to forge M_prev, but somehow whole orchestra -- existing packet fields, along w/ the Q and A -- fits in 508 bytes.
20:20 dulapbot Logged on 2021-09-19 16:28:22 asciilifeform: but only <=508byte guaranteed not to frag.
20:20 PeterL solution: don't let dr. evil be your only peer into the pestnet?
20:20 asciilifeform PeterL: he aint your peer. he's sitting somewhere in your l2+.
20:20 asciilifeform hears an l3 msg before you do, and forges it.
20:21 asciilifeform an' nao yer back to indistinguishability .
20:23 asciilifeform this was wai asciilifeform tossed out the orig. variant of the scheme. the extra packet mass does 0 good, yer back to where you started.
~ 37 minutes ~
21:00 asciilifeform ... could, of course, reveal the A's in a (dedicated broadcast type) msg ~after~ satisfied that erryone on net received yer msgs. but this brings you back to 'can't distinguish in real time'
~ 2 hours 7 minutes ~
23:08 * asciilifeform looked again for 'canhaz lamport but weighing over9000x less?' but found still only ye olde winternitz's method, which still order of magnitude too obese (and slow) .
23:08 dulapbot Logged on 2020-11-13 13:25:17 asciilifeform: meanwhile, in olds : in '19, erlehmann apparently wrote an implementation of Winternitz's OTS algo, in the style of his earlier remake of asciilifeform's Lamport OTS proggy.
← 2022-01-23 | 2022-01-25 →