ChanServ changed the topic of #freedesktop to: https://www.freedesktop.org infrastructure and online services || for questions about freedesktop.org projects, please see each project's contact || for discussions about specifications, please use https://gitlab.freedesktop.org/xdg or xdg@lists.freedesktop.org
<karolherbst>
looks like gnome deployed anubis on their gitlab instance
<karolherbst>
maybe we really should do the same to get rid of the AI scrap bots and bots in general
<karolherbst>
but users report that it took them like 100 seconds to complete the challenge, which would be too annoying 🙃
<dwfreed>
might be tunable?
<karolherbst>
yeah, should be, but I'm sure the dev would also help us out with those aspects
<karolherbst>
but yeah.. that would be a complete pain for bot networks
<karolherbst>
would be worth it if we could get rid of the spam protection we have in place
haaninjo has quit [Quit: Ex-Chat]
Xe_ has joined #freedesktop
<karolherbst>
ah yes, the dev already arrived :D
<Xe_>
hi o/
Xe_ is now known as Xe
<Xe>
it is me, the Xe
<Xe>
I hear you have a problem with scrapers
<karolherbst>
though not sure we'd deploy/use anubis in the next couple of weeks/months given admins have to migrate to new hosting and all that
<karolherbst>
but yes
<Xe>
yeah, it's a really annoying problem
<Xe>
admittedly the solution i have now is a bit of a hack
<Xe>
I've been trying to find ways to avoid PoW, but my experiments are failing because headless browsers can bypass them lol
<karolherbst>
most spam we get is for SEO, so they often create accounts with links in their bio and never touch the account ever again 🙃 but our gitlab instance is also quite slow from time to time and crawlers just make it worse
<Xe>
yeah the crawlers are a huge pain
<Xe>
most of the time you can get them to stop by doing user agent matching and then sending a 4Ki error page with HTTP status 200
<Xe>
(has to be 200, otherwise they requeue and try again later)
<karolherbst>
I don't know what's the ETA for the migration, but I think we should properly figure out why everything is slow
<karolherbst>
though at some point we had almost a million registered accounts, because spammers were creating accounts non stop 🙃
<Xe>
I mean, the main ways that they murder git forges is by clicking every link on every page, causing the forge to do more and more expensive operations to dig deeper and deeper into git history
<karolherbst>
yeah...
<Xe>
yeah, the only real way to defend is to configure your git forge to only show history to registered accounts
<karolherbst>
mhh....
<Xe>
but that like
<Xe>
sucks
<Xe>
lol
<Xe>
(and iirc gitlab doesn't let you do that? I'm pretty sure gitea doesn't)
<Xe>
the really annoying thing is that a lot of browser use agents do random pathing for mouse movement like users do
<karolherbst>
but yeah, forcing clients to do some math problem (even if it's just proving you have working JS) sounds like a good enough hack
<Xe>
i was gonna make a "drag around this in random patterns" check and then gauge the entropy of it, but browser use agents are smart enough to crack that
<Xe>
actually karolherbst the way Anubis does it is actually kinda clever
<Xe>
it doesn't just check if you have JS support, it checks if you support modern JS (many scrapers don't support ES modules), it checks for browser features that scrapers compile out, and it does a proof of work check
<karolherbst>
even better
scrumplex has joined #freedesktop
<karolherbst>
I think as long as it's not running for more then 1 or 2 seconds, pow should be fine
<Xe>
the trick for that is to set the proof of work difficulty down to 1 (8 bits)
<karolherbst>
doesn't have to be perfect, just annoying enough to get rid of some of the spammers/scrappers
<Xe>
PoW difficulty 1 usually passes instantly
<karolherbst>
would be interesting to see the impact after migration and having good enough monitoring there
<Xe>
yeah, I'm gonna try and get in touch with the GNOME people to see what impact they've had
<Xe>
just wish they didn't use matrix lol
<Xe>
may just email them
<karolherbst>
yeah, would be interesting to hear. Though not sure how big gnomes instance is compared to others
<karolherbst>
ours is... massive 🙃
<Xe>
yeah, when I saw the specs of the server you need I kinda instantly understood how big y'all really are lol
scrumplex_ has quit [Ping timeout: 480 seconds]
<karolherbst>
yeah...
<Xe>
(I'm a recovering former SRE)
<karolherbst>
our CI system is no joke 🙃
<Xe>
is gitlab still a pain in the ass to put on multiple boxes?
<Xe>
iirc last time I tried it was impossible
<karolherbst>
well...
<karolherbst>
we solved that issue
<Xe>
that sounds like there's a fun story there
<karolherbst>
mesa runs CI job across runner farms in several places
<Xe>
btw, you gonna be at scale this week?
<karolherbst>
and we do GPU testing and all that. A normal mesa pipeline has like 500 jobs 🙃
<karolherbst>
not sure...
<Xe>
I'm speaking on saturday (about AI stuff for my employer)
<karolherbst>
not really an infra person myself, I'm mostly just annoyed by spam, crawlers and our gitlab instance being slow 🙃 and kinda helping out to figure out a good solution there
<karolherbst>
anyway.. I should go to bed, it's getting late here. At least I'd be super interested in trying out anubis to see if it helps with anything, but I suspect it's gonna have to wait after the migration anyway
<Xe>
if you need help, let me know
<Xe>
given you're not an infra person, if you have trouble, it's a documentation bug
<karolherbst>
cool, thanks
<karolherbst>
I mean, I'm not doing admin stuff myself, that's for the actual admins here. Just having admin access through coc work and sometimes dealing with reports on spam myself
<Xe>
oh you work with Lyude don't you
<karolherbst>
yeah
<Xe>
been friends for years lol
<karolherbst>
so I've heard
<karolherbst>
Lyude was surprised we follow each other on socials 🙃 but I think we've had a chat about it a while ago, lol
<Xe>
the internet is much smaller than we think
<karolherbst>
anyway, I'm going to get some sleep. Will ping you once we try it out and/or run into troubles or something
alanc has quit [Remote host closed the connection]
alanc has joined #freedesktop
TrinitronX has quit [Quit: Lost terminal]
Kayden has quit [Quit: reboot to repro bugs]
digetx has quit [Ping timeout: 480 seconds]
Kayden has joined #freedesktop
ximion has quit [Remote host closed the connection]
digetx has joined #freedesktop
lsd|2 has joined #freedesktop
lsd|2 has quit []
sima has joined #freedesktop
cascardo_ has joined #freedesktop
cascardo has quit [Ping timeout: 480 seconds]
jsa1 has joined #freedesktop
tzimmermann has joined #freedesktop
sghuge has quit [Remote host closed the connection]
sghuge has joined #freedesktop
mvlad has joined #freedesktop
luc has quit []
luc has joined #freedesktop
AbleBacon has quit [Read error: Connection reset by peer]
<bentiss>
daniels: sigh... looks like using debian12 on the AX41 machines on hetzner is going to be complicated: it installs fine, reboots fine, but after an upgrade -> dead next reboot
<bentiss>
got to reset it to rescue mode :(
<bentiss>
(and no access to the console except by making an appointment with support)
<bentiss>
__tim: did you have any troubles with your hetzner servers?
<__tim>
I've had some troubles with the AX102 range of servers, they would sometimes just shut off, and I would have to file a ticket to have them power cycled to start up again. Support would then say "it was powered down" (as if I did that) and tell me to check my syslog or my windows license (these are linux machines). But there's nothing in the
<__tim>
syslogs, it just stops. Would only happen at most every few months though, so not a biggie. But those machines you can't power on via the console again, which is apparently expected. Coincidentally that's the range where they're doing mainboard replacements now and I wouldn't be surprised if that's related ;)
<__tim>
Otherwise I don't think we've had issues, and we run debian on most of them
<__tim>
oh, they do sometimes take a VERY long time to come back after reboot
<__tim>
memory integrity check or something iirc
<bentiss>
but the first boot it quite quick, so I was expecting something in the same range
<bentiss>
(I was actually surprised because the equinix ones are taking forever to boot, but I have access to the console which is way better)
<bentiss>
anyway, thanks __tim :)
<bentiss>
hmm... ubuntu 24.04 rebooted correctly and quickly
<bentiss>
MrCooper: the retry was on a different runner
<bentiss>
(ml-30 vs ml-31)
<MrCooper>
so it was a runner issue?
<bentiss>
yeah, I'm re-pulling the image. It was probably agressivelly deleted by some script
<MrCooper>
thanks
ximion has joined #freedesktop
The_Loko has joined #freedesktop
frytaped has joined #freedesktop
frytaped is now known as Guest10553
Guest10553 is now known as frytaped
AbleBacon has joined #freedesktop
dcunit3d has quit [Ping timeout: 480 seconds]
jsa1 has quit [Ping timeout: 480 seconds]
tzimmermann has quit [Quit: Leaving]
fomys_ has joined #freedesktop
jsa1 has joined #freedesktop
<bentiss>
daniels: \o/ managed to connect a bare metal hetzner machine with a k8s on the cloud VMs. That was epic. Though it seems the cloud controller doesn't let you use private IPs for that, so further work is probably required :(
gchini has quit [Remote host closed the connection]
gchini has joined #freedesktop
<xorgy>
So I'm thinking maybe I would like to make a documentation tool to replace docbook for X11 protocol docs
<xorgy>
docbook is an absolute mess
<xorgy>
And a lot of docs in the x11 sphere are not well formed because people don't build the docs, because building the docs is so painful
<xorgy>
I'm thinking html5 with some microformats to make <dl>/<dd> more structured, and to make stable relative crosslinks that don't fail the doc build if you don't have the referenced doc installed.
jsa1 has quit [Ping timeout: 480 seconds]
<xorgy>
the idea being that the docs should be readable as plain HTML in a browser, and the doc tool would ‘upgrade’ it with resolved crosslinks and a base stylesheet that styles listings, maybe highlights and lines code blocks and inline code, and makes it readable on more screen sizes (viewport meta, etc.)
<xorgy>
I am in no position to, nor would I want to, push this on anyone, so I'll just work on it and present it ‘finished’. If anyone has any wishes or thoughts on the matter feel free to share them with me. I am building the tool in Rust so that it can be built and run as a static self-contained binary.
<alanc>
docbook was originally chosen so we didn't have to maintain the tools ourselves, but could use the same tools many other projects were already using
<alanc>
and to get us onto a single format, instead of the previous mix of troff, framemaker, and others
<xorgy>
Yeah, I have watched the rise and fall of docbook over my lifetime. I remember being frustrated with it when I was 12, I'm 28 now and I'm still frustrated with it... I understand the arguments that were made for it, but most manuals that have been imported into docbook from roff and others were never carefully revised, or even revised at all, after they were imported.
<xorgy>
there's now docbook 5 which is presented as a pile of manuals and XSLTs, and no popular or straightforward tooling
<alanc>
I'm sure there are better choices today than there were 15+ years ago when we made the choices, but the constraint of not having people to maintain custom tooling just for us is much worse today than it was back then
<xorgy>
absolutely
<alanc>
and we've certainly seen other projects migrate off docbook in the meantime
<alanc>
a lot of the X.Org docs in docbook are specifications that shouldn't change that much, not general user docs
<xorgy>
yup
<xorgy>
though a lot of them are very sparse, and the barriers to entry for building the docs mean that people don't touch them
<xorgy>
even when they absolutely should be touched (my XIM experience for example)
<alanc>
but yeah, as the person who has done most of the posting of the formatted output to the website in past years, it would be nice to have something better, since I've not even bothered to tackle that for quite a few years now
<xorgy>
my experience with going into XIM was that the documentation (imported from troff a couple decades ago, with no proofreading) oscillates between useless and misleading... and probably people much more knowledgeable of XIM have looked at it and thought to themselves "well I would clarify that, but I don't know if my changes will be valid/buildable"
fomys_ has quit []
Kayden has quit [Quit: -> JF]
DodoGTA has quit [Quit: DodoGTA]
DodoGTA has joined #freedesktop
sima has quit [Ping timeout: 480 seconds]
sima has joined #freedesktop
<Lyude>
oh yes I know Xe
<Lyude>
hi Xe :)
<Xe>
hi Lyude
a_fantom has quit [Ping timeout: 480 seconds]
Kayden has joined #freedesktop
mvlad has quit [Remote host closed the connection]
The_Loko has quit [Quit: Leaving]
sima has quit [Ping timeout: 480 seconds]
aswar002_ has quit []
grinder has joined #freedesktop
grinder has quit [Remote host closed the connection]