Post Akv7fvOGJ8UcJ0tkG0 by PJB@mastodon.gamedev.place
(DIR) More posts by PJB@mastodon.gamedev.place
(DIR) Post #Akv7fvOGJ8UcJ0tkG0 by PJB@mastodon.gamedev.place
2024-06-23T14:00:42Z
6 likes, 6 repeats
Just spent 30 minutes making this
(DIR) Post #AkvPWtRpxTHmqmPhL6 by bartholin@fops.cloud
2024-08-13T13:04:01.523565Z
1 likes, 1 repeats
@PJB wait, packets order the wrong are in the
(DIR) Post #Akw3AZEGRna7wDyRH6 by icedquinn@blob.cat
2024-08-13T20:28:09.241322Z
3 likes, 2 repeats
i saw a game once that broadcast everything it did like >ten times in UDP because "that made it more reliable"
(DIR) Post #Al6eMtu72lffQ6tKy0 by puppygirlhornypost@transfem.social
2024-08-13T21:11:20.946Z
1 likes, 0 repeats
@PJB@mastodon.gamedev.place ok but I like quic
(DIR) Post #Al6eOwR6bGvfqvISsy by lispi314@udongein.xyz
2024-08-18T23:11:58.405723Z
0 likes, 0 repeats
@icedquinn Sounds like they didn't feel like digging into TCP tuning.
(DIR) Post #Al6eOwyma3rrXN3MNU by icedquinn@blob.cat
2024-08-18T23:12:29.116344Z
1 likes, 1 repeats
@lispi314 TCP is hopeless.
(DIR) Post #Al6hngk0OJdUHEd3MO by lispi314@udongein.xyz
2024-08-18T23:19:46.657357Z
0 likes, 0 repeats
@icedquinn Probably why QUIC was made, yes.
(DIR) Post #Al6hnhU5cyV0aABrBw by icedquinn@blob.cat
2024-08-18T23:50:33.816242Z
0 likes, 0 repeats
@lispi314 QUIC is not bad. I don't like the monotonically increasing counters† but otherwise I would probably use it.Wanted to implement it for nim before i burned out.†i dislike counters like this because they could theoretically run out. once they do, what then? just crash the program? everyone just tells me "don't worry about running out of int64s"
(DIR) Post #Al6lOsUtwZqrJdzO0u by gentoobro@shitpost.cloud
2024-08-19T00:30:54.572556Z
0 likes, 0 repeats
You'd have to go really fast for a long time to run out of int64's. You'll run out of unix timestamps in a similar order or time, so it's not much different from a timestamp.
(DIR) Post #Al6pUToj172YKFBjJQ by lispi314@udongein.xyz
2024-08-19T00:53:48.785568Z
0 likes, 0 repeats
@gentoobro @icedquinn For similar reasons it annoys me that unix timestamps aren't bignums.
(DIR) Post #Al6pUUjRcEPVA9j1ii by rheaplex@social.xenofem.me
2024-08-19T01:07:29.132389Z
0 likes, 0 repeats
@lispi314 @icedquinn @gentoobro 1969 to 2038 was a loooooong way off.
(DIR) Post #Al6pUVXmb4fzgHHEBM by gentoobro@shitpost.cloud
2024-08-19T01:16:43.463193Z
0 likes, 0 repeats
2^64 is a hell of a lot bigger than 2^32. "Longer than the age of the universe by multiple orders of magnitude" bigger.
(DIR) Post #Al6qUVQx6fFZIFe3DU by lispi314@udongein.xyz
2024-08-19T01:26:14.962772Z
0 likes, 0 repeats
@gentoobro @rheaplex @icedquinn And just about a single bitflip from causing a crash if the error isn't properly handled.
(DIR) Post #Al6qUWSPIk0YT3Kj7g by gentoobro@shitpost.cloud
2024-08-19T01:27:57.184355Z
1 likes, 0 repeats
ECC ram. If you're flipping bits in ram randomly, you're fucked. Timestamps are the least of the problems. All your pointers are now segfaults waiting to happen.
(DIR) Post #Al6qWUtw3cWCDl9Qy8 by icedquinn@blob.cat
2024-08-19T01:28:18.141923Z
0 likes, 0 repeats
@rheaplex @gentoobro @lispi314 variable length identifiers are what CRDTs use now. it does complicate some stuff. old formats were built in the age of everything-is-a-db-record.in the case of network protocols a "i ran out of handles—negociate restart" is fine. what happens now is basically undefined behavior (likely it will just wrap because thats what the machine code does, but you are supposed to track those so if it wraps back to 1 it would be interpreted as trying to send outdated packets)