Show Idle (>14 d.) Chans
← 2022-01-04 | 2022-01-06 → ↓
00:21 awt hm yeah might only be checking that keys are unique per peer
00:22 awt also fwiw probably only checking the first key for a peer
~ 1 hours 11 minutes ~
01:33 asciilifeform per spec a peer can have as many keys as operator wants; the most recently validated one is used by default for transmission
01:34 asciilifeform ( and when rekeying, that's the one that gets rekeyed , but iirc we aint there yet )
01:34 asciilifeform *validating
01:35 * asciilifeform thinking, possibly oughta change spec there, use ~randomly selected~ key when there are multiples
01:35 jonsykkel http://logs.bitdash.io/pest/2022-01-04#1001584 << which bot is this? sourc avialbable? so i can fix if is problem with my pestron
01:35 bitbot Logged on 2022-01-04 17:33:23 shinohai[billymg]: bot won't connect to smalpest in c AT ALL so I mite retire this bot
01:36 asciilifeform jonsykkel: not sure whether shinohai posted the src to his bot
01:36 asciilifeform it's the 1 that spits exch rate etc
01:37 jonsykkel i see. only tested irc connection from irssi and weechat so far
01:37 asciilifeform it worked fine on fleanode and currently working on dulapnet; would hate to see it vanish, if its incompat. w/ a pestron, it's the latter that's buggy
01:38 jonsykkel probably. my "ircd" is 100lines or so. maybe something is missing
01:39 asciilifeform could be; not tried it w/ mine yet
01:39 asciilifeform iirc billymg did not need major mods to make his go
01:39 asciilifeform ( iirc his bot is a largely unmodified clone of mine )
01:39 jonsykkel would help to see src to find out wats going on either way
01:40 asciilifeform aha
01:42 asciilifeform jonsykkel: http://paste.deedbot.org/?id=tjM8 btw
01:43 jonsykkel asciilifeform: peered
01:43 asciilifeform a++ !
01:43 asciilifeform that was quick
01:44 jonsykkel fastest peerer in the west
01:44 jonsykkel getting msgs directly from u now
01:44 asciilifeform record, to date
01:44 jonsykkel gotta set up leaderboard
01:47 jonsykkel also re: this - it does autorekey. and >1key tested also
01:47 bitbot Logged on 2022-01-04 21:48:26 asciilifeform[billymg]: (we dun have autorekeying yet, afaik, except possibly in 'smalpest')
~ 52 minutes ~
02:39 asciilifeform jonsykkel: oh neato
~ 27 minutes ~
03:07 signpost http://logs.bitdash.io/pest/2022-01-04#1001606 << I'll cut ya a patch if you want, but figured it was a one liner and best rolled into your next batch of changes.
03:07 bitbot Logged on 2022-01-04 19:08:59 awt[billymg]: 9983 is the latest available patch. no one released a patch for what signpost tried.
03:07 signpost but happy to if you like.
03:09 * signpost a little slow atm, house still on the mend, but getting better.
03:09 signpost I have ftr had flus that were 10x this.
↖
03:13 whaack awt: perhaps there should be a notice that gives an error for when you try to direct message someone you don't have a key for, just tried to dm you to confirm it would not work (since we are not peered) and had to check logs to see that indeed my msg to you failed.
↖
03:13 signpost whaack: do we need to roll keys on our peering?
03:15 whaack signpost: i understand your question as "should we make a new key pair (because *i* sent out the same key to multiple people) and if that is the correct understanding of your q the answer is no, since i used the key you gpg'd to me.
↖
03:17 signpost ah right, sorry, groggy yet.
03:17 whaack also awt ^ the msg 'awt: perhaps there should be..." passed my dupe filter, i sent the msg, and then saw my own msg along with a relay from signpost
03:18 signpost whaack: yeah, I saw a dupe of that message from asciilifeform.
03:19 whaack i'm going to send the same msg again to see if it somehow the problem is specific to the text, ignore the next line:
03:19 whaack awt: perhaps there should be a notice that gives an error for when you try to direct message someone you don't have a key for, just tried to dm you to confirm it would not work (since we are not peered) and had to check logs to see that indeed my msg to you failed.
03:19 whaack ^ nope, this time the same msg had no dupes
03:20 signpost iirc awt said he had some bug which sometimes mutates messages such that they evade the dedup cache.
↖
~ 41 minutes ~
04:01 awt If anyone happens to spot the issue in the code I would welcome a patch
04:01 awt these dupes are annoying as hell
04:04 asciilifeform http://logs.bitdash.io/pest/2022-01-05#1001700 << asciilifeform's latest bout doesn't stand up to any comparison with flu. was much moar subtle, low-level annoyance, but protracted, 2wk long fever haze
04:04 bitbot Logged on 2022-01-05 03:09:11 signpost[billymg]: I have ftr had flus that were 10x this.
04:05 asciilifeform http://logs.bitdash.io/pest/2022-01-05#1001701 << per spec oughta show eggog, aha
04:05 bitbot Logged on 2022-01-05 03:13:02 whaack[billymg]: awt: perhaps there should be a notice that gives an error for when you try to direct message someone you don't have a key for, just tried to dm you to confirm it would not work (since we are not peered) and had to check logs to see that indeed my msg to you failed.
04:06 asciilifeform http://logs.bitdash.io/pest/2022-01-05#1001703 << pestron ought have balked when you tried to make a 2nd peer w/ key seen prior
04:06 bitbot Logged on 2022-01-05 03:15:42 whaack[billymg]: signpost: i understand your question as "should we make a new key pair (because *i* sent out the same key to multiple people) and if that is the correct understanding of your q the answer is no, since i used the key you gpg'd to me.
04:06 asciilifeform (i.e. is buggy if it dun do this)
04:07 asciilifeform http://logs.bitdash.io/pest/2022-01-05#1001710 << there's defo a bug atm which foils deduping, aha
04:07 bitbot Logged on 2022-01-05 03:20:28 signpost[billymg]: iirc awt said he had some bug which sometimes mutates messages such that they evade the dedup cache.
04:07 * asciilifeform not had chance to hunt for it just yet
04:08 whaack i'm hunting down the bug, as a way to familiarize myself with the code as well
04:11 asciilifeform possibly a lead, is the fact that iirc all the dupes i've seen on the current blatta were 'from self'
04:11 asciilifeform (e.g. asciilifeform[awt]: foo)
04:12 asciilifeform btw eggogs of the type where e.g. sent to a keyless peer prolly oughta emit 'wallops' so to be visible in whatever pane of irctron
04:12 * asciilifeform puts this on list of what oughta go in spec
04:14 asciilifeform in 'modal' irctrons, e.g. 'weechat', the usual kind tends to go unnoticed
04:15 * asciilifeform currently pesting via 'weechat', tho for fleanode/dulapnet normally uses xchat
04:15 asciilifeform (xchat is ill-behaved when multiple servers)
04:15 jonsykkel ftr had another issue peering with blatta cuz dedup check happens after message is logged (infosec.py line238 and line258)
04:16 asciilifeform tsk
04:16 jonsykkel dunno if related to bug discussed
04:16 * asciilifeform must bbl
04:17 jonsykkel (bouncebakd messages from self got logged twice in db and caused problems)
04:19 jonsykkel (se also http://zzz.st/up/jT70pm4A/9983-tmp-disgusting-hack.vpatch)
04:25 whaack i can see in my db log that the first mesage has a half byte (nibble) that gets set to a 0 in the following messages, which is why i imagine that the dupe is not detected
04:30 whaack the original mesage had an odd number of nibbles, i.e. oddlength hexstring for bytes, so atm looks life self messages are erroneously corrupt for some reason with this half-byte suffix, and the relays correct this, which then results in a duplicate
~ 10 hours 44 minutes ~
15:14 whaack greetings to pestnet
~ 24 minutes ~
15:38 awt whaack: greetings
15:40 awt whaack: glad to hear you're checking out the source
15:46 whaack awt: it's a pleasure reading, going through the full blatta category on your blog atm
15:48 whaack as i mentioned in the above lines, it appears that somehow a half-byte is getting written ocassionally to the body of the message, or maybe i just saw 0x20 as a halfbyte because it terminates with a 0
15:49 whaack in any case the irc client -> local server line has this extra 0x20, which gets eliminated on the relays, causing a dupe message
15:50 whaack very possibly the bug is on the end of certain irc clients, which for whatever reason send some junk byte to blatta
15:52 awt whaack: hmm but the dedup check isn't until after the message has been decrypted
15:53 awt or hmm
16:00 whaack awt: right, my understanding of the problem is somewhere between locasl irc client and blatta an extra junk byte gets padded on, in my one example i've found the junk byte is 0x20, or maybe it's just an extra 0x2, then during the whole encrypt, send over wire, decrypt, this junk byte gets removed, and thus inside the db th
16:04 awt oh ok so you're saying junk is being added on to the unecrypted message bytes before they are stored in the db and then when the message comes back those bytes are gone?
16:07 whaack awt: precisely
16:10 asciilifeform awt: potentially a dump cmd that'd shit out hexdump of last N msgs might be handy in case like this
16:12 asciilifeform (i.e. you see a dupe, can get a dump and see whether mutation hypothesis holds)
~ 1 hours 13 minutes ~
17:26 whaack i'm going to spam for a bit, to check if i can reproduce the dupe message bug, please ignore the below
17:26 whaack trying to reproduce dupe, please ignore 1
17:26 whaack trying to reproduce dupe, please ignore 2
17:26 whaack trying to reproduce dupe, please ignore 3
17:26 whaack trying to reproduce dupe, please ignore 4
17:26 whaack trying to reproduce dupe, please ignore 5
17:27 whaack trying to reproduce dupe, please ignore n
17:27 whaack trying to reproduce dupe, please ignore n
17:27 whaack trying to reproduce dupe, please ignore n
17:27 whaack trying to reproduce dupe, please ignore n
17:27 whaack trying to reproduce dupe, please ignore n
17:27 whaack done, no dupes :/
17:30 whaack going to try again, this time with a trailing space
17:30 whaack ^ i think that is the bug, trailing spaces are getting chomped out
17:30 whaack yup, 0x20 is a space
17:31 asciilifeform interesting btw failure of embargo queue:
17:31 asciilifeform UnicodeDecodeError: 'ascii' codec can't decode byte 0xe2 in position 111: ordinal not in range(128)
17:31 asciilifeform ^ lol, crashed when pasting
17:32 asciilifeform http://paste.deedbot.org/?id=3MhE << the paste
17:32 whaack at first i threw 0x20 into an ascii converter in order to see what character it was, and i got no visual response (since the character is a space " "), so i assumed it was a junk non-ascii byte, which delayed my discovery of the bug
17:45 awt testing trailing space
17:45 awt motherfucker
17:46 whaack solution is to make sure that relays don't remove the trailing space, imo
17:46 awt nice work whaack
17:47 awt lulz
17:47 whaack ty
17:47 awt tab completion in weechat adds a trailing space
17:52 awt also seen for the first time in the wild: whaack[4]
17:52 whaack nice
17:52 whaack might as well leave ourselves unpeered for testing purposes
17:55 whaack also asciilifeform: if i may add a suggestion to the spec, i think that if you receive a message from the same peer who sent the message, there should be a way to also include all those who relayed to you
17:55 whaack atm the fact that you are peered to the origin means you lose information in your console
17:55 whaack err, ircclient
17:57 whaack so say you are peered to me, and i send you a msg, you may want to receive <whaack[*|awt|signpost], where the * symbolizes you received the msg from me
17:57 whaack and if there are many peers then <whaack[*|5]>, for example
~ 1 hours 16 minutes ~
19:14 signpost here's some whitespace
19:14 signpost heh!
19:14 signpost good find whaack
19:14 signpost ^ I don't think that one had trailing whitespace btw
19:14 signpost but I might've accidentally hit the spacebar
19:17 awt signpost: tab completion adds a space
~ 50 minutes ~
20:08 asciilifeform whaack: not sure i get it -- why wouldja want to mark relayers if you in fact got a nonhearsay (immediate) msg ?
20:08 asciilifeform the immediate copy is the 'authoritative' one
20:08 asciilifeform if you have it, dun really matter who else gives a copy
20:09 asciilifeform the [] notation is currently to mark hearsays, such that is clear via whom the hearsay came
20:11 signpost aaa
20:12 signpost re: tab completion
20:12 whaack i guess just to see the net topology, but perhaps not necessary for every message
20:12 signpost probably best to get that info from logs
20:12 signpost more room for additional detail
20:12 whaack or a command
20:13 asciilifeform oh ha just got a signpost and a signpost[4] of same msg
20:14 asciilifeform (either embargo mechanism is buggy, or immed. copy actually delayed by >1s?)
20:15 asciilifeform whaack: maybe i'm thick, but as i see it 'i got the immediate and these-here relay copies' doesn't add useful info
20:15 asciilifeform if you have the immediate copy, you know the message is authentic (unless the peer's key was compromised somehow)
20:16 whaack right, it's to know the health of your topology for every msg, but not very useful
20:17 whaack asciilifeform: the original msg i believe has the whitespace, and all the relays do not
20:17 asciilifeform the [] info is useful in 2 cases : 1) you aint in fact peered with the originator. in that case you want to know via which peer he is able to reach you 2) the immed. copy was lost in transit somehow. then useful to see that N of yer peers in fact had received it
20:17 asciilifeform whaack: if the whitespace mutates, this is a bug
20:17 asciilifeform messages are to be unmolested in relay.
20:18 whaack yes, that is the bug, somehow in the encrypt/decrypt/relay process trailing whitespaces are getting nixed
20:18 asciilifeform right makes sense
20:18 whaack i'm looking for where exactly it's happening, have not found yet
20:20 * signpost perhaps late to the game, but in unrelated lulz, vitalik is now being forthright about ethereum centralization.
20:20 signpost https://archive.is/U1Xfj
20:20 asciilifeform 'yes it's a paypal, kthx' ?
20:20 signpost mmmhm
20:21 asciilifeform 'censorship is still prevented' lol
20:21 signpost "inevitable" etc, but hey, block *validation* still "decentralized" lol
20:21 whaack poor guy needs to use some of his 'billions' to get himself some food
20:22 asciilifeform ever since the dao thing, the paypalism was obv. to all but perhaps down's syndrome sufferers
20:26 asciilifeform whaack: d00d swallowed, what, 100k btc. would be surprised if he's short of food.
20:26 asciilifeform even if he only gets to eat 1% and handlers eat the rest
20:28 signpost https://archive.is/wxGgt fishwraps cranked up
20:28 signpost "Finally, there???s one more big winner I see coming this year: regulators."
20:29 * signpost looks forward to covid passport smart contract on ethereum determining whether he can purchase petrofood.
20:32 asciilifeform doesn't seem readily distinguishable from past N yrs of reichpress ethertardium agitprop
20:32 whaack idk, perhaps the stress of running such a high profile scam is what has taken such a toll on his body, he does not look healthy
20:33 asciilifeform whaack: prolly mp's disease (substance abuse)
20:33 asciilifeform the grim reaper of noveau-riche errywhere
20:35 asciilifeform 'why ~shouldn't~ i party 24/7', crash, rinse, repeat, 'doctor's orders', switch $substance a coupla times, croak.
20:48 whaack test space
↖
20:53 whaack ^ signpost or anyone else, any chance you *didnt* see a dupe of that msg?
~ 15 minutes ~
21:08 billymg vitalik has always looked like that though, i think it's part of his "brand"
21:10 billymg saw him described somewhere as "vitalik is what a dumb person thinks a genius looks like" -- i.e. turbo autist too busy being glued to his screen creating The Next Big Thing to bother to eat
↖
21:12 billymg whaack: fwiw did not see a dupe of http://logs.bitdash.io/pest/2022-01-05#1001834
21:12 bitbot Logged on 2022-01-05 20:48:11 whaack[billymg]: test space
21:13 whaack billymg: makes sense, we are not peered, so only the relays were sent to you, and all of the relays remove the whitespace
21:16 billymg ahh, i see. makes sense now
21:31 asciilifeform whaack: i saw dupe: whaack | test space whaack[signpost] | test space
21:35 signpost whaack: no dupe over here, just one 'test space' msg
21:36 signpost trim's probably the wrong thing here eh? should just consider the message to be w/e was sent
21:37 asciilifeform signpost: aha, per spec there oughtn't to be any molestation of the msg
21:37 asciilifeform i.e. whatever is received from irc frontend, after filtering commands, oughta get sent
21:37 asciilifeform (ditto when relaying)
21:39 asciilifeform http://logs.bitdash.io/pest/2022-01-05#1001837 << as one'd expect, when usg.mit did casting call, they picked the expected sort of trained seal, as for any other hollywood production
21:39 bitbot Logged on 2022-01-05 21:10:20 billymg: saw him described somewhere as "vitalik is what a dumb person thinks a genius looks like" -- i.e. turbo autist too busy being glued to his screen creating The Next Big Thing to bother to eat
21:49 whaack i found the problem, pretty sure
21:50 whaack irc clients append \r\n to the end of every msg, awt is using python's rstrip() method to get rid of that, but the function rstrip also gets rid of trailing whitespaces
21:51 whaack awt: offending line is 244 in infosec.py
21:51 signpost yeah, there'll be no end to the special casing that kind of thing unless it's removed entirely.
21:54 awt interesting ty whaack
21:54 whaack np
~ 1 hours 56 minutes ~
23:50 asciilifeform ohey anuther: awt | interesting ty whaack awt[signpost|whaack] | interesting ty whaack
← 2022-01-04 | 2022-01-06 → ↑

Random(pest) | Download hourly DB snapshot | Get Source Code

Theme: Light | Dark | System Default