| 03:04 |
signpost |
got that lisp encoder on par with the python impl. seems like there's plenty of optimization to be had yet. |
| 03:04 |
signpost |
unnecessary recomputation of the alias method lookup table was bogging it down. |
| 03:05 |
signpost |
asciilifeform: phf: any notions re: efficient xor of subsequences of a given bit vector? seems like using displaced arrays nukes optimization of bit-xor |
| 03:06 |
signpost |
http://paste.deedbot.org/?id=z1oZ << current sauce |
| 03:07 |
signpost |
http://paste.deedbot.org/?id=kAql << and tests |
| |
↖ |
| 03:07 |
signpost |
barring something smarter, I might just chop the thing into many non-displaced bit-vectors instead of trying to do everything in one giant bit-vector |
| 03:09 |
signpost |
since later this thing will need to eat/shit files on disk, the in-memory mega-array will have to go anyway. who says what's to be broadcast/received fits in RAM. was just a place to start. |
| 03:09 |
* |
signpost bed, night all |
| |
~ 7 hours 33 minutes ~ |
| 10:42 |
asciilifeform |
signpost: pretty neat. you'll want file chunks as direct msg tho, rather than broadcast, if you let'em all-to-all no amt of bw will ever suffice. iirc we had a thread where 'let each station offer operator-selected files to own peers'. |
| |
↖ |
| 10:44 |
* |
asciilifeform atm has this in the new spec draft in anticipation of signpost's thing |
| 10:46 |
asciilifeform |
asciilifeform's notion is, can issue 'i want file with $hash' to >1 peer, and get direct lubyisms from >1 if they have it |
| |
~ 45 minutes ~ |
| 11:31 |
signpost |
yep that sounds right. I was thinking it might be nice to inform peers how much spray you want from each to conserve bandwidth also, just as a politeness mechanism. |
| 11:33 |
signpost |
"do you have $hash?" -> "yep, and I call it foo, and it's $size" -> "pls canhaz 20mb of encoded fileblox" |
| 11:33 |
signpost |
the "do you have" can be used to get a sense of availability of $hash in your net |
| |
~ 15 minutes ~ |
| 11:49 |
asciilifeform |
signpost: 'how much firehose' is actually a tricky q, esp. if one considers peers w/ wildly mismatched pipes |
| 11:49 |
asciilifeform |
prolly oughta indicate in the request. 'i can eat n pieces / sec.' |
| 11:49 |
signpost |
that's true. I'm thinking we don't want each peer to piss 1gb at receiver if file's 1gb though |
| 11:50 |
asciilifeform |
signpost: for that, oughta cut into blocks (say 1MB), w/ indiv. hashes |
| 11:51 |
asciilifeform |
in general, oughta limit the amt of piss triggerable by 1 packet to sumthing reasonable |
| 11:51 |
asciilifeform |
otherwise, too easy to get folx to drown ~selves~ in own piss, and not even through malice but by sheer accident |
| 11:53 |
signpost |
yeah, perhaps something like you say. I initiate a transfer from peer p1 telling him "spray me packets at up to $limit" and p1 decides (or is limited by pipe) where beneath the limit the transfer operates (or above, nothing hard governs it obviously) |
| 11:53 |
signpost |
if I'm not saturated on my end, I can ask P2, P3, etc for same, incrementally filling up the amount of my pipe I wish to dedicate to the transfer |
| 11:54 |
asciilifeform |
in practice, most folx w/ pestron will be limited by cpu cost of decoding, rather than bw per se |
| |
~ 2 hours 12 minutes ~ |
| 14:07 |
unpx |
Just in case someone was interested, I just recalled nsucrypto exists and that there are some people trying to figure out the solution of this unsolved puzzle: https://nsucrypto.nsu.ru/archive/2022/round/2/task/7/#data |
| 14:08 |
unpx |
Ah. Silly. Here's the complete task list https://nsucrypto.nsu.ru/media/Olympiads/2022/Round_2/Tasks/22_second_round-kdjcf.pdf |
| 14:08 |
* |
asciilifeform had an ancient orcish b00k as a kid, w/ puzzlers quite like this |
| |
~ 6 hours 9 minutes ~ |
| 20:18 |
asciilifeform |
unpx: problem 9 (where asked to find a symmetric cipher that is commutative and can be used in shamir's 3pass algo) iirc provably has no solution (or at any rate not such that anyone would want) |
| 20:18 |
* |
asciilifeform would be curious to learn otherwise |
| 20:19 |
* |
asciilifeform wouldn't mind having one for pest rekey but afaik aint no such thing w/out bignumism |