Hide Idle (>14 d.) Chans


← 2022-09-17 | 2022-09-19 →
awt[asciilifeform]: jonsykkel: Would it be possible to decrease the frequency of ignore messages smalpest sends out? Currently seeing 1 per second. I'm wondering if blatta may not be able to handle that rate currently due to the shitty pure python serpent implementation.
jonsykkel[asciilifeform]: awt: sure, set it to 10sec(avg) now
awt[asciilifeform]: ty jonsykkel
asciilifeform: !q uptime
dulapbot: asciilifeform: time since my last reconnect : 0d 0h 8m
asciilifeform: lol ^ wat
asciilifeform: ^ embargoed for ~week?!
asciilifeform: !q uptime
dulapbot: asciilifeform: time since my last reconnect : 0d 0h 9m
awt[asciilifeform]: blatta currently consuming about 3% CPU on my laptop - some kind of ryzen
phf: ayo holup
awt[asciilifeform]: Hey phf was testing from behind a firewall on another IP for a while this morning
phf[asciilifeform]: another test
phf[asciilifeform]: y'all network is funky
phf[asciilifeform]: yeah, but the context is #a, wherein we discuss the odd and strange delays on pest network
awt[asciilifeform]: I'm on 9972, yes, no prod loops, no cpu issues on this machine.
dulapbot: (asciilifeform) 2022-09-18 asciilifeform: a and still slow?
awt[asciilifeform]: yeah been following
awt[asciilifeform]: Got to separate out potential issues.
phf[asciilifeform]: i think i'm back to being near realtime, but the chain that was broken due to previous test messages been sitting in limbo for ~~5 minutes
phf[asciilifeform]: http://logs.bitdash.io/pest/2022-09-18#1013045 << also when you sent this message, did you just switch over the ip/client, or you've been on current setup for a while?
bitbot[asciilifeform]: Logged on 2022-09-18 12:48:52 awt: Hey phf was testing from behind a firewall on another IP for a while this morning
awt[asciilifeform]: in blatta the default order buffer expiration is 180s
awt[asciilifeform]: shamefully for a couple minutes this morning I was operating from 2 ips
awt[asciilifeform]: forgot to shutdown the station on my laptop before starting up on the server
asciilifeform: that'll do interesting things to chain
phf[asciilifeform]: well, i've updated my AT quite a while ago, and was sending "test transmit through awt". in fact things didn't start working until awt sent the "Hey phf" message above
awt[asciilifeform]: that was immediately after I came back online at my old IP
phf[asciilifeform]: well, then all the mysteries are solved. awt was down, my ATs were all out of date, asciilifeform's advertised ip is actually behind nat, dulapbot has always been magically and mysteriously delayed. i think that covers it
asciilifeform: ( if there were a working fork detector, oughta've gone off errywhere )
asciilifeform: ( or at least errywhere from whose pov is hearsay )
awt[asciilifeform]: lol yes. I can't seem to connect to asciilifeform from behind a nat either.
asciilifeform thinking about it, doesn't recall ever actually turning the embargo knob on dulapbot down from 180
asciilifeform on local rack recently installed a new pfsense box and oughta take anuther look at the config, strongly suspects the fwd rule aint actually wurking
phf[asciilifeform]: i suspect we might need stations in variety of configurations. one useful thing to have would be a station without operator, that acts more like a backbone: no retransmit delays, doesn't do its own chain discovery, mostly concerned with maintaininig healthy AT, etc.
asciilifeform suspects that once erryone has working addrcast, and has at least 1 peer that doesn't habitually wander away ( e.g. the racked bots ) the 'came back from beach and ugh can't connect to anyone' headache will largely evaporate
phf[asciilifeform]: http://logs.bitdash.io/pest/2022-09-17#1013014 << it was not like the typical perturbator output, that seem to be default soundtrack in hackerspaces and nerd get togethers. i think the health&petrurbator combo are supposed to be some kind of noise/industrial/witchhouse/salem thing. it was less synthwave and more i dunno in
phf[asciilifeform]: dustrial rock? the sound system was also pretty bad. i very much enjoyed the experience because i was just flying round in the moshpit, which was also at the center of the acoustic noise cone. my friends here were sitting down thought it mostly sucked.
bitbot[asciilifeform]: Logged on 2022-09-17 01:13:11 signpost: http://logs.bitdash.io/pest/2022-09-16#1012988 << ah I like that song Excess by Perturbator. believe I've heard the other band but can't recall.
phf[asciilifeform]: the general theme was a kind of rehash of the early 2000s: "i'm a piece of shit trash, so lets just shoot heroin and die, idk"
phf[asciilifeform]: but reinacted by middle class nerds, who bought fresh hot topic shit for the occassion :} it was a very polite mosh, without much bottoming out. you could kind of see people going "ow too much too much i'm not feeling comfortable about all this, but i'm going to act cool somewhere else"
asciilifeform: phf: wat's 'bottoming out' ?
phf[asciilifeform]: asciilifeform, it's when you surrender to the experience, without much concern for physical consequences
phf[asciilifeform]: like a fat dude killing a can of beer, smashing it on his head, rushing the mosh, wiping out, accidentally getting his teeth kicked out while on the ground, then getting up while grinning through blood, and rushing the mosh again
phf[asciilifeform]: i miss a good mosh ._.
phf[asciilifeform]: http://logs.bitdash.io/pest/2022-09-17#1013018 << i generally don't like indie riffs, even with country folk affectation. it reminds me of late 90s early 2000s american tv show thing, where they would have some guest band play at the local in-show bar. the one tree hill weekly guest band.
bitbot[asciilifeform]: Logged on 2022-09-17 01:15:39 signpost: incidentally just got in from seeing the mountain goats, was great, if you're into that earnest folk music dork thing, which I am.
phf[asciilifeform]: i went to an appalachian fest yesterday, had some pretty nice lineup, and i like that among that traditional bluegrass bands that's been coming there for years you also have a bunch of hipster bands doing a kind of medley of appalachian/irish/scottish covers
dulapbot: Logged on 2022-09-18 12:58:12 asciilifeform: that'll do interesting things to chain
bitbot[asciilifeform]: Logged on 2022-09-18 12:59:09 asciilifeform[jonsykkel|awt|crawlerbot]: ( if there were a working fork detector, oughta've gone off errywhere )
asciilifeform: ^ likely not the only 1
asciilifeform: this 1 esp. odd considering that asciilifeform is one of the few folx directly peered to it
asciilifeform gearing up to make yer anuther attempt to load the embargo logic in curr. blatta into head
asciilifeform looking fwd to publication of phf's cl proto
awt[asciilifeform]: Anyone interested in trying a blatta that uses mcrypt serpent?
awt[asciilifeform]: Now testing mcrypt...
signpost[asciilifeform]: awt: happy to test w/e you got.
awt[asciilifeform]: Only thing preventing me from doing a release right now is I'm still seeing some cold peers being updated to my own ip.
phf[asciilifeform]: what's mcrypt?
awt[asciilifeform]: phf: more serious crypto lib recommended by authors of pure python serpent currently in use: http://mcrypt.sourceforge.net/
phf[asciilifeform]: also one perhaps not kosher but pragmatic solution to "crypto in python slow" is to cffi to openssl :>
phf[asciilifeform]: or cffi to a c lib of some sort
phf[asciilifeform]: ah mcrypt looks like that very "c lib of some sort"
awt[asciilifeform]: phf: this is how I'm using mcrypt, cffi
phf[asciilifeform]: cool. i'll take a look at it also, because i've been considering doing an ffi to openssl. perhaps not a permanent solution, but a stop gap before i extract and optimize relevant code from ironclad
awt[asciilifeform]: Actually, the latest blatta disables ac by default, so the self-address bug can be avoided for now at least by simply not using AC. Can at least test out potential "prod loop" fix and see if mcrypt speeds up blatta on lower powered machines.
awt[asciilifeform]: Ah hmm last instance of the self address bug could have been due to having multiple stations using the same keys at different ips.
asciilifeform: phf: openssl incls serpent?!
phf[asciilifeform]: asciilifeform, no. (i just discovered that, the thought was in the back of my mind, rather than explored option)
asciilifeform: could ffi classic c serpent naturally
asciilifeform: a loox like awt baked this already
asciilifeform orthogonally: thinking re when is the correct time, hypothetically, for station to hammer ports. imho oughta be when received addrcast, and knows that atm peer is live, but attempt to connect to the ephemeral port in the addrcast not resulted in war
dulapbot: (asciilifeform) 2022-09-18 asciilifeform: we'll need this, eventually, asciilifeform suspects, for 'final solution to nat'
asciilifeform: ... not resulted in warm peer within some interval
asciilifeform: hrm thought the msg procrustation was fixed in 9973 lol
asciilifeform: thing is, the hammer gotta work from ~both~ ends
phf[asciilifeform]: yeah, could combine classic c serpent, and sha256/512 from busybox into a lib, ffi into it, cffi from python, etc.
asciilifeform: ( as asciilifeform understands, it will do so tho )
asciilifeform: ( each station will receive addrcast from the other; then $interval elapses w/out connection in the usual way; then each starts hammer and continues until connects )
dulapbot: Logged on 2022-09-07 11:00:58 asciilifeform: afaik with 'symmetric' nats of the type apparently victimizing jonsykkel , the only drill that worx is to hammer random ports, from both directions, until match
awt[asciilifeform]: message chunking fix was 9972
asciilifeform must bbl
awt[asciilifeform]: [9971 with mcrypt, potential prod loop fix, ac disabled by default][http://share.alethepedia.com/blatta/]
awt[asciilifeform]: Due to flow of liquishit, won't have much time to debug/support tomorrow - fri.
awt[asciilifeform]: Also, if you don't have mcrypt built, should still "magically" just work using pure python. It will emit warnings in the log, however.
signpost[asciilifeform]: awt: cool, will give that a try. will note any problems here, but don't take it as a demand to stop what you're doing and address immed.
signpost[asciilifeform]: awt: fwiw the try...except blocks without specific errors to catch are going to swallow bugs and cause you pain later
signpost[asciilifeform]: oughta probably do the pivot on whether mcrypt is available on an ImportError handler
signpost[asciilifeform]: would you mind linking the mcrypt you recommend using?
awt[asciilifeform]: signpost: yes. should be a link in the readme.
signpost[asciilifeform]: ah derp, should've read
awt[asciilifeform]: Still seeing up to 45% cpu on out outgoing ac bursts. Gotta smooth that out.
phf[asciilifeform]: what's ac?
signpost[asciilifeform]: lemme make sure that isn't my own src manglement, gonna re-press
awt[asciilifeform]: odd. I added in the interpret_none_as_zero method for that case.
signpost[asciilifeform]: yeah I'd hand-mangled the file already, my fault
signpost[asciilifeform]: (to fix that bug)
awt[asciilifeform]: phf: ac is address cast
awt[asciilifeform]: ok what I'm seeing now is deedbot repeatedly prodding my station
signpost[asciilifeform]: crash loop, sec
signpost[asciilifeform]: (unwanted touching!!!)
awt[asciilifeform]: my theory was that prod_interval_seconds was being interpreted incorrectly somehow and causing a prod every time the event loop runs.
signpost[asciilifeform]: ok, up now with less mangled src
signpost[asciilifeform]: (deedbot is up)
awt[asciilifeform]: nice no prod loop now
awt[asciilifeform]: looks to be syncing for some reason
signpost[asciilifeform]: yeah, was doing some getdata, now not sure, no output yet
awt[asciilifeform]: embargoed for default period of 100+ secs likely
signpost[asciilifeform]: cpu usage is like 5%, nb!
signpost[asciilifeform]: got embargo at 1sec already in knobs
phf[asciilifeform]: like watching water boil :)
signpost[asciilifeform] pokes deedbot with a stick
signpost[asciilifeform]: his blatta saw that msg immediately
awt[asciilifeform]: pretty snappy now
signpost[asciilifeform]: yep seems fine
signpost[asciilifeform]: thanks for the new release awt!
awt[asciilifeform]: signpost: yw - thanks for testing
phf[asciilifeform]: well, it's fairly obvious that embargo has an effect, but what i don't i understand. so packet goes into embargo, getdata for rest of chain comes out, at that point there should be enough packets to release all the embargoed stuff, why does it still hang?
awt[asciilifeform]: phf: the algo is messages get dumped from the embargo not when the ancestor comes in, but simply after embargo period expires.
awt[asciilifeform]: this is wrong(tm)
awt[asciilifeform]: Was optimized for first sync, assuming a large backlog
signpost[asciilifeform]: alrighty my blatta's also up to date.
phf[asciilifeform]: come play with us
awt[asciilifeform]: Ah just thought of another way you could end up with self-addressed peers
awt[asciilifeform]: You blast out ac, shut down your station, then immediately start it up.
awt[asciilifeform]: Since blatta only keeps ac hashes in memory, any of your ac messages that might come back to you will not be detected as dupes after you restart.
asciilifeform: http://logs.bitdash.io/pest/2022-09-18#1013174 << thing is, it seems to happen to immediate msgs (e.g. asciilifeform -> dulapbot ) , and not only to hearsays
bitbot[asciilifeform]: Logged on 2022-09-18 21:04:07 awt: phf: the algo is messages get dumped from the embargo not when the ancestor comes in, but simply after embargo period expires.
asciilifeform: per spec, immediates only get embargoed if mischained
asciilifeform: but seems to happen erry single time anyffin (incl. immed from asciilifeform) hits dulapbot
asciilifeform: or hm, looking at debug log, is consistent w/ the picture where hearsay hits it, and gets in the chain b/w 2 asciilifeform -> dulapbot immed msgs. so not mystery, nm
asciilifeform: ( and yea the order buffer oughta evict when either timeout or ancestor is found , this will be in 0xfa spec )
dulapbot: Logged on 2022-06-20 14:35:41 asciilifeform: ... the 'order buffer' otoh would be superfluous in a 'mutable' pestron.
asciilifeform wrote spec strongly around the assumption that frontend is irclike and therefore must attempt to squeeze output to console in sumthing maximally approaching a logical order
bitbot[asciilifeform]: Logged on 2022-09-18 20:15:50 awt: [9971 with mcrypt, potential prod loop fix, ac disabled by default][http://share.alethepedia.com/blatta/]
asciilifeform must bbl
awt[asciilifeform]: $ticker btc usd
← 2022-09-17 | 2022-09-19 →