Newest at the top
| 2025-02-22 00:32:04 +0100 | caconym | (~caconym@user/caconym) (Quit: bye) |
| 2025-02-22 00:25:42 +0100 | peterbecich | (~Thunderbi@syn-047-229-123-186.res.spectrum.com) (Ping timeout: 252 seconds) |
| 2025-02-22 00:24:29 +0100 | <haskellbridge> | <sm> see you |
| 2025-02-22 00:23:21 +0100 | <haskellbridge> | <magic_rb> Anyway, sleep time |
| 2025-02-22 00:21:40 +0100 | <haskellbridge> | <magic_rb> Im planning to try to shoehorn vodozemac into ement.el (emacs matrix client) to get e2ee |
| 2025-02-22 00:21:30 +0100 | noxp | (~ensyde@2600:381:df29:528d:5c12:4661:6233:7eba) (Ping timeout: 268 seconds) |
| 2025-02-22 00:21:08 +0100 | <haskellbridge> | <magic_rb> And for e2ee youll have to bind against vodozemac which is in rust |
| 2025-02-22 00:20:44 +0100 | <haskellbridge> | <sm> oh interesting, that's good news |
| 2025-02-22 00:20:31 +0100 | <haskellbridge> | <magic_rb> Crypto is largely a client side thing only, https is ofc server side, but e2ee is completely client side |
| 2025-02-22 00:20:26 +0100 | <haskellbridge> | <sm> To be honest, with bridges out of the picture I rarely see state resolution / split brain issues on matrix. (I see it more often on irc, or used to) |
| 2025-02-22 00:19:07 +0100 | <haskellbridge> | <sm> And another part must be the amount of data to be managed, though you could restrict that |
| 2025-02-22 00:18:44 +0100 | ensyde | (~ensyde@70-33-158-82.unassigned.ntelos.net) (Ping timeout: 252 seconds) |
| 2025-02-22 00:18:35 +0100 | <haskellbridge> | <sm> I imagine another hard part of building a matrix server is the cryptography, eg availability of the right good libraries. |
| 2025-02-22 00:18:18 +0100 | <haskellbridge> | <sm> Yes.. I look forward to v3. There's a nifty state visualiser now which must be a help |
| 2025-02-22 00:17:46 +0100 | <haskellbridge> | <magic_rb> And other fun things, dunno much about the actual impl, just what i heard while hanging out in matrix dev adjacent rooms |
| 2025-02-22 00:17:13 +0100 | ensyde_ | (~ensyde@70-33-158-82.unassigned.ntelos.net) |
| 2025-02-22 00:16:43 +0100 | <haskellbridge> | <magic_rb> Effect is that people might not see each others messages. You might get rejoined into a room you left, because the leave event disappears |
| 2025-02-22 00:16:15 +0100 | <haskellbridge> | <magic_rb> If you hear "split brain" its a stateres bug, where room state permanently diverges and different servers see and advertise different state. |
| 2025-02-22 00:15:40 +0100 | <haskellbridge> | <magic_rb> And that part is the buggiest part of matrix afaik |
| 2025-02-22 00:15:26 +0100 | <haskellbridge> | <magic_rb> Then after they all rejoin, stateres happens where it has to merge again |
| 2025-02-22 00:15:19 +0100 | <haskellbridge> | <sm> I hear state resolution algorithm v3 is coming |
| 2025-02-22 00:15:10 +0100 | <haskellbridge> | <magic_rb> Say you have 8 HSs in a room, 4 split off for 4 days but the users in them can still communicate between themselves |
| 2025-02-22 00:15:05 +0100 | <haskellbridge> | <sm> tricky algorithms should be our bread and butter |
| 2025-02-22 00:15:04 +0100 | <geekosaur> | byzantine |
| 2025-02-22 00:14:40 +0100 | noxp | (~ensyde@2600:381:df29:528d:5c12:4661:6233:7eba) noxp |
| 2025-02-22 00:14:35 +0100 | <haskellbridge> | <magic_rb> The state res is the worst, as in, the way possibly divergent event graphs are merged |
| 2025-02-22 00:13:49 +0100 | <haskellbridge> | <sm> geekosaur: Complex ? |
| 2025-02-22 00:12:41 +0100 | ensyde | (~ensyde@70-33-158-82.unassigned.ntelos.net) ensyde |
| 2025-02-22 00:12:39 +0100 | <haskellbridge> | <magic_rb> Im gonna focus on my haskell filesystem for now, then maybe something else in haskell |
| 2025-02-22 00:12:31 +0100 | weary-traveler | (~user@user/user363627) user363627 |
| 2025-02-22 00:11:53 +0100 | <haskellbridge> | <sm> everyone struggles to write a matrix server it seems, haskell should be among the best languages for the writing/maintaining part at least (not necessarily the efficiency/operating at scale part) |
| 2025-02-22 00:11:29 +0100 | <haskellbridge> | <magic_rb> Pretty sure haskell is fast enough. The best way to speed up synapse which is in python, is to speed up the database, postgres |
| 2025-02-22 00:10:52 +0100 | <haskellbridge> | <magic_rb> sm if you decide to try to run your own HS, whichever you choose. Feel free to ping me for help |
| 2025-02-22 00:10:38 +0100 | <haskellbridge> | <sm> it could be a good reference / research implementation. In future, auto translate it to rust or something for better efficiency |
| 2025-02-22 00:10:07 +0100 | <geekosaur> | Matris protocol is nuts |
| 2025-02-22 00:10:03 +0100 | <haskellbridge> | <magic_rb> Saying to run one ig, sorry, i see RIIH everywhere these days lol |
| 2025-02-22 00:09:40 +0100 | <haskellbridge> | <magic_rb> Ah |
| 2025-02-22 00:09:39 +0100 | <geekosaur> | in Haskell or any other language |
| 2025-02-22 00:09:33 +0100 | <geekosaur> | oh, misunderstood. no I'm not interested in writing one |
| 2025-02-22 00:09:30 +0100 | <haskellbridge> | <magic_rb> As in, i dont think there is a tangible benefit to using haskell for a HS, most of the complexity lies in implementing the whole spec, the stateres algorithm and actually making the internal data model fast |
| 2025-02-22 00:09:17 +0100 | merijn | (~merijn@host-vr.cgnat-g.v4.dfn.nl) merijn |
| 2025-02-22 00:09:12 +0100 | atwm | (~andrew@19-193-28-81.ftth.cust.kwaoo.net) (Ping timeout: 252 seconds) |
| 2025-02-22 00:08:19 +0100 | <haskellbridge> | <magic_rb> And tbh i dont think haskell would solve much |
| 2025-02-22 00:08:06 +0100 | <haskellbridge> | <magic_rb> Writing a matrix HS is insanely difficult |
| 2025-02-22 00:07:50 +0100 | <haskellbridge> | <magic_rb> And no you dont need as much power as i have, especially if you dont run synapse. But even with it im giving it too much memory, its possible to tune and run on just 4GB of memory, i just threw more ram at it because i couldnt be arsed to tune it |
| 2025-02-22 00:07:49 +0100 | <geekosaur> | I've offered a couple of times but HF seems uninterested |
| 2025-02-22 00:06:54 +0100 | <haskellbridge> | <sm> when do we get the rock solid haskell homeserver... |
| 2025-02-22 00:06:46 +0100 | <haskellbridge> | <magic_rb> Im running synapse because i have the hardware and its the most reliable and drama free. Conduit got forked at least twice over drama |
| 2025-02-22 00:06:14 +0100 | <haskellbridge> | <magic_rb> Synapse i in python, dendrite in go, conduit, conduwuit and grapevine are forks of each other, all in rust |
| 2025-02-22 00:05:50 +0100 | <haskellbridge> | <magic_rb> But your main bottleneck if you do run this wont be the bridges but rather the matrix homeserver. You have 5 options, synapse - heavy, reliable, dendrite - light, broken beyond belief, conduit - light, slow progress, conduwuit - light, possibly toxic maintainer, grapevine - light and i dunno |