| norayr | so it worked. unfortunately not the second option. but the first option, used after the second option didn't work. | 09:02 |
|---|---|---|
| norayr | but i was searching something for this vainly. | 09:02 |
| norayr | thank you for the link. | 09:03 |
| poetaster | Thaodan, ah, I had overlooked that... | 10:14 |
| poetaster | Thaodan, https://github.com/poetaster/fahrplan/releases/tag/2.0.41 | 11:18 |
| poetaster | lbt, fatal: unable to access 'https://github.com/poetaster/fahrplan/': Could not resolve host: github.com | 11:42 |
| poetaster | piggz[m], rinigus maybe someone sees something I've overlooked: https://github.com/poetaster/fahrplan/releases/tag/2.0.41 | 11:46 |
| cos^ | Just came to check this room out. There's a larger sailfish room at #saiflishos:matrix.org , i think it would be worth merging the rooms at one point. Having two rooms on same topic is a bit confusing. | 13:07 |
| poetaster | matrix clients suck universally. | 13:32 |
| cos^ | not relevant on this issue | 13:36 |
| Thaodan | cos^: To where is the room bridged so far? There's also a liberachat room which is currently only used to tell users to go to oftc. | 13:38 |
| Thaodan | poetaster: Yes they suck.. I prefer IRCv3 every day so far. | 13:38 |
| Thaodan | Haven't found a good one for emacs for example. | 13:38 |
| cos^ | Thaodan: it's not bridged anywhere | 13:39 |
| Thaodan | ok | 13:39 |
| poetaster | Thaodan, I have/had weechat running matrix, but it's limited. I'm constantly writing scripts to administer synapse. nightmare. | 13:39 |
| poetaster | Thaodan, worst abuse of postgres I have ever witnessed. | 13:40 |
| Thaodan | poetaster: I know of that plugin, I tried to use it to connect an irc client to weechat but sadly it didn't work. | 13:40 |
| poetaster | Thaodan, took a while, but I have one colleague that's still using it (work). I use gomuks for work. | 13:41 |
| Thaodan | postgres is very good, sad so hear its getting abused. | 13:41 |
| poetaster | Thaodan, yes. I mean, it's quite common for dbs to be used incorrectly. still. | 13:41 |
| Thaodan | It's sad that matrix-ircd or purple-matrix isn't going anywhere. | 13:41 |
| Thaodan | poetaster: Yes I think so, I still read the frequent discussions on akonadi's uses of dbs. | 13:42 |
| cos^ | For those still wishing to use IRC clients, there is https://github.com/progval/matrix2051 | 13:42 |
| Thaodan | cos^: thanks | 13:43 |
| poetaster | cos^, the room in question is on the central matrix.org server? | 13:43 |
| poetaster | cos^, https://medium.com/@RiotChat/new-irc-integrations-oftc-and-snoonet-b88883a58303 | 13:44 |
| cos^ | poetaster: rooms are not on any specific server, they are decentralized between all participating ones | 13:44 |
| poetaster | cos^, not on my matrix server. I don't federate. | 13:44 |
| Thaodan | cos^: sounds very, very nice however that reminds be I wanted to improve IRCv3 in emacs-circe and Communi-Sailfish.. | 13:44 |
| Thaodan | I need a 48h day.. | 13:44 |
| cos^ | poetaster: there are about 30 servers in that room, so it's distributed on all of them | 13:45 |
| poetaster | cos^, as far as I can tell, you can bridge the oftc rooms into that room with no real work. | 13:46 |
| poetaster | cos^, maybe try #_oftc_#somewhere:matrix.org | 13:46 |
| poetaster | cos^, newer docs https://matrix-org.github.io/matrix-appservice-irc/latest/bridged_networks.html | 13:47 |
| cos^ | poetaster: i don't think it makes much sense to bridge to IRC, it's not relevant anymore. Also major networks require kicking users after idle timeout which sucks. | 13:47 |
| poetaster | cos^, 'relevant' to whom. Jeez. Might as well use whatsapp | 13:48 |
| Thaodan | I agree it is subjective. | 13:48 |
| poetaster | I know the internals of the matrix synapse reference server. the only working one. it is a fucking disaster. | 13:48 |
| cos^ | Dendrite and conduit are not working ones? | 13:49 |
| poetaster | dendrite is not production ready. conduit I'm not familiar with. | 13:49 |
| poetaster | in any case, I'm here for technical work and not chat. sorry. | 13:50 |
| poetaster | for that matter, synapse is not production ready. I have fail2ban rules to keep people from scrapping media urls. | 13:52 |
| Thaodan | 😓 | 13:54 |
| Thaodan | That sounds bad | 13:54 |
| poetaster | Thaodan, I have paranoid users. | 13:54 |
| cos^ | Do you have issue link for it? | 13:54 |
| Nico | scraping media urls? | 13:55 |
| Nico | Does that mean link previews? | 13:55 |
| cos^ | First time I hear of the issue, I'd like to hear more info on it | 13:56 |
| poetaster | It's a well known issue. It's mitigated by media being encryption, numerous tickets: https://github.com/matrix-org/synapse/issues/1403 | 13:56 |
| Nico | Well, accessing media currently is unauthenticated, since it is hosted below a random id | 13:56 |
| poetaster | shit, shouldn't have mentioned it. we've been here before. | 13:56 |
| Nico | But you can use that patch to add authentication to the media repo: https://github.com/deepbluev7/synapse/tree/media-repo-auth | 13:56 |
| Nico | (Still writing up the spec change proposal to enforce that network wide) | 13:57 |
| poetaster | Nico, we discussed this some time ago. I implemented a log watcher that logs ips without user sessions. those ips are banned if they access anything | 13:57 |
| poetaster | ah, cool | 13:57 |
| cos^ | Um, the URLs have a long random part that acts as "key" to prevent unauthenticated access. Regular rate limiting prevents scraping for them. | 13:57 |
| Nico | cos, still, adding auth to those endpoints is not a bad idea at this point (since web clients can now add auth headers to image tags) | 13:58 |
| Nico | Especially since some people don't understand that key part and it allows revoking access | 13:58 |
| poetaster | 'don't understand' ah, ha. | 13:59 |
| cos^ | Nico: in some situations yes.. but that would break sharing the links outside Matrix clients (for example links to IRC channels) | 13:59 |
| poetaster | this is the wrong forum for this discussion. | 13:59 |
| Nico | cos, yes, but such usecases should be opt in imo | 13:59 |
| poetaster | Nico, thanks for the patch. | 13:59 |
| Nico | (Which is why my patch has an option to allow that) | 13:59 |
| Nico | poetaster, just be aware that the stricter settings break clients like Element :3 | 14:00 |
| cos^ | Yeah, but claiming " synapse is not production ready" because of this issue is just plain FUD | 14:00 |
| poetaster | cos^, have you looked at how synapse uses postgres? | 14:01 |
| Nico | I have heard so many people claim that, that it is probably just a good idea to change it, so that people need to look for something else :D | 14:01 |
| cos^ | poetaster: no. | 14:01 |
| poetaster | cos^, sec, I'll give you a glimpse. | 14:02 |
| poetaster | Nico, who what the genius that decided timestamp should be in milliseconds insteaf of unix epoch? | 14:02 |
| cos^ | poetaster: but claiming that software that is used by >50M users including large companies, governments and public services to be not production ready isn't factually correct. | 14:03 |
| Nico | poetaster, Why would you want second accuracy for a high bandwidth messaging service? | 14:03 |
| Nico | That sounds terrible | 14:03 |
| Nico | Arguably milliseconds is low resolution | 14:03 |
| Thaodan | Is unix epoch more memory consuming? | 14:04 |
| poetaster | No, less. | 14:04 |
| poetaster | It's just a quibble. | 14:04 |
| Nico | It is exactly the same on most systems. Most systems use 64bit for both | 14:04 |
| Nico | (You don't store a unix timestamp in a 32 bit int.) | 14:05 |
| poetaster | Nico, yes, I'm not quibbling with the resolution per say, just how it's stored. there is a native way to save the higher resolution. | 14:06 |
| cos^ | poetaster: I don't care about how it uses postgres. Traditionally it's been quite memory hog but that has been improving constantly. Memory is no longer such huge issue for server admins. | 14:06 |
| poetaster | cos^, as as system administer responsible for a matrix server, I'm constantly appalled. | 14:07 |
| Nico | poetaster, in most cases it is stored as a string inside of the events json, so it doesn't really make a difference how you store it | 14:07 |
| Nico | Especially since state snapshots are 50 to 95% of the db | 14:07 |
| poetaster | Nico, you just made my point. dumping serialized data into a column. | 14:08 |
| Nico | (95% if you don't run the space compressor) | 14:08 |
| poetaster | thanks for the patch, it's in the ticket system: 41150efe5a4e4dc6d02e01c07f416949e93481a0 | 14:08 |
| Nico | Just make sure you test if it works for you before deploying | 14:10 |
| poetaster | Nico, a question, in theory you can set retention policies at a room level? | 14:11 |
| Nico | Since it DOES break stuff | 14:11 |
| Nico | poetaster, yes, but that feature is experimental and partially broken still | 14:11 |
| Nico | Needs some MSC rework | 14:11 |
| poetaster | Nico, thought as much. broke for me :) | 14:11 |
| poetaster | So I wrote scripts using purge. which of course need custom time calculations. | 14:12 |
| poetaster | Nico, is there another production ready server out there? | 14:12 |
| Nico | Probably the sane option for now | 14:12 |
| Nico | No, but dendrite is pretty close already | 14:12 |
| poetaster | Ok, thanks. | 14:13 |
| Nico | (Has the same limitations on media storage though, since that is required by the spec atm) | 14:13 |
| cos^ | AFAIK Dendrite's lacking push notifications but otherwise it's completely usable. Not sure about SSO and module support though, they might be lacking. | 14:13 |
| poetaster | Thaodan, you have any idea why obs would claim: fatal: unable to access 'https://github.com/poetaster/fahrplan/': Could not resolve host: github.com | 14:15 |
| Nico | Hm, push notifications should work now, they were incorrectly sending timestamps as milliseconds instead of seconds | 14:15 |
| poetaster | Nico, man you just keep making my arguments for me :) store epoch + offset, and don't dump those into json arrays because that forces you to parse ... blah, blah. | 14:16 |
| Nico | No, the problem is that push notifications sent to the gateway for some reason use epoch ;p | 14:17 |
| poetaster | Nico, from the dendrite server? | 14:18 |
| Nico | Well, the gateway needs epoch, but dendrite sent matrix timestamps :3 | 14:18 |
| poetaster | Ok, that's what I meant. Epoch is obviously not enough, in my mqtt broker expermiments I stored 2 fields. | 14:22 |
| poetaster | Nico, but I'm writing some stuff using clock_gettime at the moment since, unlike date, it supports down to nanoseconds. | 14:27 |
| poetaster | ah, in time.h there is also gettimeofday with support. | 14:29 |
| poetaster | I'm being thick: date -u +'%Y-%m-%dT%H:%M:%S.%3NZ' | 14:30 |
| poetaster | %3 is milliseconds. | 14:31 |
| poetaster | rabbit hole. iso 8601 to millisecond resolution: date -u +'%Y-%m-%dT%H:%M:%S.%3N%Z' | 14:32 |
| poetaster | in my matrix postgres scripts I do: select to_timestamp(received_ts/1000)::date from events limit 5 | 14:38 |
| poetaster | which is ok for mose purposes. | 14:39 |
| poetaster | I'd store %Y-%m-%dT and %H:%M:%S.%3N in separate fields. the first field will take up practically no space. | 14:40 |
| Nico | Sounds useless ;p | 14:41 |
| poetaster | I like to complicate things :) nah, used something like this in an extension to an mqtt broker to push extra time headers to subscribers. | 14:42 |
| poetaster | if I remember correctly, it was pretty useless :) | 14:42 |
| poetaster | Now I've spammed here quite enough. goes shopping. | 14:45 |
| poetaster | piggz, rinigus I'm no longer able to update packages in chum:testing. | 17:03 |
| poetaster | piggz, rinigus get stuff like: Error while triggering rebuild for sailfishos:chum:testing/fahrplan: No permission to modify project 'sailfishos:chum:testing' for user 'poetaster'. | 17:03 |
| piggz | let me see.. | 17:03 |
| piggz | youre still a maintaner in fahrplan | 17:04 |
| piggz | problem for lbt :D | 17:04 |
| poetaster | I thought maybe I screwed something up. | 17:04 |
| poetaster | fatal: ambiguous argument '2.0.41': unknown revision or path not in the working tree. looks odd, but it correct across release, tag, spec | 17:05 |
| poetaster | I'll ping lbt next week. | 17:05 |
| poetaster | swiss users of chum will be so pissed :) | 17:06 |
| poetaster | anyone able to build for harmattan? I can't find my old meego sdk virtual | 17:30 |
| rinigus | poetaster: tried to replace with commit ID for fahrplan. then got an error: fatal: unable to access 'https://github.com/poetaster/fahrplan/': Could not resolve host: github.com | 17:59 |
| poetaster | yeah. me too. I think it really is an lbt? | 17:59 |
| rinigus | poetaster: weird. that doesn't happen to my project (tested in one devel repo) | 18:02 |
| rinigus | will try to replicate my _service | 18:02 |
| rinigus | poetaster: not much better: https://build.sailfishos.org/package/show/home:rinigus:maps:devel/test# | 18:08 |
| rinigus | lbt or Keto: please look at https://build.sailfishos.org/package/show/sailfishos:chum:testing/fahrplan# | 18:09 |
| poetaster | rinigus, thanks! | 18:16 |
| poetaster | rinigus, at some time we could talk privately about an 'open mapping' business/ngo. I feel a bit out of my depth, but I have a lot of experience with startups. | 18:17 |
| poetaster | and not for profits, for that matter. | 18:17 |
| rinigus | poetaster: no profits desired over that :) | 18:31 |
| rinigus | but sure, we should do that. not now, though :) | 18:32 |
| rinigus | and it would be great to have such expertise on board | 18:32 |
| rinigus | thanks for the help! | 18:33 |
| piggz | rinigus: ive also got a github issue on OBS, cane clone a repo | 18:51 |
| Thaodan | what also helps is someone that can politely express problems | 19:32 |
| Thaodan | often that is more important than an direct knowledge | 19:32 |
| rinigus | piggz: so, it is global then... | 19:42 |
| piggz | rinigus: yeah id say so, been pinging lbt for a few days :D | 19:51 |
| Mister_Magister | noooo i locked myself out via dod packages | 22:01 |
| Mister_Magister | wheres lbt when you need him | 22:02 |
Generated by irclog2html.py 2.17.1 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!