 | Canvas Meta | 240 Members |
| General discussion for Canvas - https://canvas.toast.ooo - Bridged to Discord | 42 Servers |
| Canvas Meta | 240 Members |
| General discussion for Canvas - https://canvas.toast.ooo - Bridged to Discord | 42 Servers |
| Sender | Message | Time |
|---|---|---|
| 16 Jul 2024 | ||
 Marius | nevermind, lemmy-ui is dead anyway | 08:28:11 |
 taaz (biglemmowski.win) | I am playing with the idea of doing the Atlas as an excuse to learn svelte and svelte-kit a bit more... :D | 09:54:28 |
| Marius | The atlas use a json files to store the data, in a pretty simple format, so it should be pretty UI-agnostic (after maybe a conversion phase). To be honest, I also had the idea to rewrite the canvas server in Rust with all the cool performance that imply, but ultimatly, I’m pretty sure the most costly part is keeping the socket connection open. The rest can just be optimised. | 10:02:59 |
| taaz (biglemmowski.win) | though I guess most work would be with html5 canvas (highlighting the given art) anyway so the chosen framework is probably not that important | 10:03:31 |
| taaz (biglemmowski.win) | * though I guess most work would be with the html5 canvas (highlighting the given art) anyway so the chosen framework is probably not that important | 10:03:45 |
| taaz (biglemmowski.win) | * though I guess most work would be with the html5 canvas (highlighting the given art and so on) anyway so the chosen framework is probably not that important | 10:03:52 |
| taaz (biglemmowski.win) | In reply to @marius851000:mariusdavid.frafaik the most costly thing is opening the underlying websocket, keeping it up open is almost free | 10:04:45 |
| taaz (biglemmowski.win) | In reply to @marius851000:mariusdavid.fr* afaik the most costly thing is opening the underlying websocket, keeping it open is almost free | 10:04:55 |
| taaz (biglemmowski.win) | obviously the more chatty each client conn is the heavier it gets | 10:05:17 |
| Marius | I’ve also the potential idea, if performance ever become a problem, of having an authoritative server only communicating to child servers, and then it’s those servers that server data to user, with the only communication between parent/child being canvas updates, log-in and user placing/undoing pixels | 10:05:31 |
| Marius | * I’ve also the potential idea, if performance ever become a problem, of having an authoritative server only communicating to child servers, and then it’s those servers that serve data to user, with the only communication between parent/child being canvas updates, log-in and user placing/undoing pixels | 10:05:49 |
 soda_cans | In reply to @marius851000:mariusdavid.frjust with js you can get a very fast pixel server, ive been on one that could handle like 10,000 pixels a seconds, most limiting factor was the browser not liking that to render | 10:07:11 |
| soda_cans | but that might also have been just run on a very good server | 10:07:27 |
| soda_cans | * but that might also have been run on a very good server | 10:07:35 |
| soda_cans | In reply to @marius851000:mariusdavid.fr* just with js you can get a very fast pixel server, ive been on one that could handle like 10,000 pixels a second, most limiting factor was the browser not liking that to render | 10:07:43 |
| Marius | I should not forget JS have some proper multithreading. Unlike Python. | 10:08:16 |
| Marius | I think? | 10:08:48 |
| taaz (biglemmowski.win) | In reply to @marius851000:mariusdavid.frthat will hopefully change in the upcoming years ! py3.13 has big stuff coming for it in terms of the GIL and threading perf | 10:09:14 |
| taaz (biglemmowski.win) | https://labs.quansight.org/blog/free-threaded-python-rollout | 10:10:19 |
| taaz (biglemmowski.win) | still experimental for 3.13 but the goal post is set for projects to start experimenting | 10:11:57 |
| Marius | this https://www.techempower.com/benchmarks/#hw=ph&test=fortune§ion=data-r22 indeed show JS can be fast for serving HTTP and database access. Of course, serving http is just a single part of the performance concern. But actually, after the cache fix, has performance been a concern? | 10:13:34 |
| Marius | * this https://www.techempower.com/benchmarks/#hw=ph&test=fortune§ion=data-r22 indeed show JS can be fast for serving HTTP and database access. Of course, serving http is just a single part of the performance concern. But actually, after the cache fix, has performance been a concern? _the library I tend to use is actually slower by a few percent of the faster Javascript library* | 10:14:45 |
| Marius | * this https://www.techempower.com/benchmarks/#hw=ph&test=fortune§ion=data-r22 indeed show JS can be fast for serving HTTP and database access. Of course, serving http is just a single part of the performance concern. But actually, after the cache fix, has performance been a concern? the library I tend to use is actually slower by a few percent of the faster Javascript library | 10:14:51 |
 @nudel:nudeltopia.de left the room. | 11:41:36 | |
 @autumnspark1226:matrix.org left the room. | 12:54:23 | |
| Marius changed their display name from Burrito (Marius) to Marius. | 14:21:22 | |
 @shittykopper:envs.net left the room. | 18:14:33 | |
 grant | socket.io can easily handle 10k concurrent connections | 18:26:28 |
| taaz (biglemmowski.win) | heyo grant, would it be at all possible to get a database dump of the pixels? Considering the log files are not absolutely exact | 18:53:58 |
| grant | In reply to @wereii:matrix.org i'm going to be regenerating the log files using the database eventually (i don't have much time today) before releasing any of the db stuff i have to verify the clean-ness of it, but i might be able to do that 🙏 | 19:01:25 |