05:15 |
awt |
asciilifeform: Not saying you were running more than one station on one lan. Just - this is exactly what happens when you sniff your own ac farts, which is not happening on my station running the exact same version. |
05:18 |
awt |
And the only way I know this can happen is if a station that didn't send them originally has the same keys - because the ac message hashes are deduped. |
05:21 |
awt |
Just saying it might be something a little more subtle, since, works here, than 'catastrophically broken' |
| |
~ 5 hours 3 minutes ~ |
10:25 |
asciilifeform |
awt: asciilifeform not fully dug into src, but 'are deduped' here means that blatta does somewhere have 'i sent this addrcast, if it comes back to me, ignore it' logic ? |
10:26 |
asciilifeform |
awt: fwiw yest. the thing sent a rather large pile of addrcasts (was stuck sending'em continuously, for yet-to-be-discovered reason, until commented out) so may have triggered some otherwise improbable bugola |
10:36 |
billymg |
http://logs.bitdash.io/pest/2023-02-23#1024089 << i have not been keeping 100% up-to-date with the recent discussion around pest/blatta. awt: which version do you recommend i press/run for bitbot? |
10:36 |
bitbot |
Logged on 2023-02-23 13:44:13 phf[awt|deedbot]: billymg, asciilifeform can you guys update your respective loggers with awt's recent multiline patch? |
10:50 |
awt |
asciilifeform: yes indeed there is such logic. Hash is added to the long buffer and long buffer is checked for incomeing ac messages. |
10:57 |
awt |
billymg: need to think about that for a bit. |
11:09 |
asciilifeform |
oh hm awt , iirc longbuffer not persists to disk ? then makes sense that restarting might cause the thing to breathe in own addrcast (assuming that it not checks speaker) |
11:11 |
asciilifeform |
( even per old draft, speaker field oughta be checked, but this apparently not made it into the proggy yet? ) |
11:11 |
asciilifeform |
'The Speaker field of a Message bearing an Address Cast Payload must match a WOT handle corresponding to its apparent originator.' |
11:24 |
awt |
asciilifeform: longbuffer in blatta is persisted to disk. |
11:29 |
awt |
asciilifeform: *actually* looks like only direct and broadcast messages make it from memory into the db - ac message hashes stored in memory only. |
11:31 |
awt |
So yes, restarting quickly definitely could cause uptake ac broadcasts originating at your own station. |
| |
~ 20 minutes ~ |
11:51 |
phf |
fancy |
11:51 |
phf |
"message type KEYOFFER jonsykkel" |
| |
~ 17 minutes ~ |
12:08 |
phf |
looks simple enough, but i need to implement persistence first. i'm procrastinating on it, because there are no good ways of doing it |
12:11 |
phf |
most convenient way to do persistence in common lisp is allegrocache, which is franz's proprietary thing. add a metaclass to a class, and its instances get magically attached to store. all the corner case work. |
12:15 |
asciilifeform |
phf: ever tried 'elephant' ? ( asciilifeform did, >decade ago, was a 1st class bitch to build but did work ) |
12:20 |
phf |
if i ever need to design a document store for pension fund managment i might reach for something like elephant |
| |
↖ |
12:23 |
* |
asciilifeform would like a commonlisp with mmap'd image, periodically wonders wai not already baked somewhere long ago |
12:27 |
phf |
mmap'd image will be in a consistent state only some part of the time |
12:27 |
phf |
because of ephemera, lack of transactional semantics, in flight data state, etc. |
12:36 |
phf |
i haven't seen a satisfying solution to this problem, probably closest is the way richard hipp writes fossil, which c backed by sqlite, but not in the way most people do it. he's got all kinds of convenience functions so that most runtime operations happen in sqlite, and c is just |
12:36 |
phf |
used to glue chunks of sqlite behaviors together. it is still as awkward as any form of language-sql hybrid |
12:37 |
asciilifeform |
imho anyffin short of the ultimate iron solution will always have same unsatisfying 'x backed by y' flavour |
| |
~ 4 hours 58 minutes ~ |
17:36 |
awt |
asciilifeform: perhaps at some point you could let me set up a station behind your nat. Would speed up debugging quite a bit. |
| |
~ 1 hours 36 minutes ~ |
19:12 |
asciilifeform |
awt: would have to get fwding working then, lol |
19:12 |
asciilifeform |
(presently it aint) |
19:13 |
* |
asciilifeform baking own pestron, but at his current pace, protons will fucking decay before finished |
| |
~ 1 hours 11 minutes ~ |
20:24 |
phf |
ascii making own pest "i have figured out a superior iron forging method, trying to work out its practical details" |
20:28 |
phf |
ascii making own pest "setun-70's base 3 architecture proved superior, but i can't fit in an ice40 unencumbered fpga!" |
20:30 |
phf |
ascii making own pest "euclid has led us fundamentally astray, some of the late 19th century alternative mathematical models might prove better foundation for my gui library" |
| |
~ 45 minutes ~ |
21:15 |
asciilifeform |
phf: lol; but sadly nope, just mired in seemingly-eternal liquishit atm |
21:18 |
asciilifeform |
http://logs.bitdash.io/pest/2023-02-24#1024262 << thinking about this, wouldn't this be the very example of when to 'allegrocache' etc ? |
| |
↖ |
21:18 |
bitbot |
Logged on 2023-02-24 12:20:22 phf[jonsykkel|awt|deedbot]: if i ever need to design a document store for pension fund managment i might reach for something like elephant |