2026/06/28

Newest at the top

2026-06-28 14:59:19 +0000 <spew> I use source hut
2026-06-28 14:54:34 +0000 <jaror> Yeah, so the choices seem to be either Codeberg or GitHub. The latter seems more likely to survive better in the current landscape, but I'd hope codeberg can survive too.
2026-06-28 14:53:00 +0000spew(~spew@user/spew) spew
2026-06-28 14:47:55 +0000vanishingideal(~vanishing@user/vanishingideal) vanishingideal
2026-06-28 14:46:10 +0000chromoblob(~chromoblo@user/chromob1ot1c) chromoblob\0
2026-06-28 14:46:00 +0000fp(~Thunderbi@46-133-26-225.mobile.vf-ua.net) (Ping timeout: 245 seconds)
2026-06-28 14:45:45 +0000chromoblob(~chromoblo@user/chromob1ot1c) (Read error: Connection reset by peer)
2026-06-28 14:43:43 +0000 <tomsmeding> so if that's the case and codeberg is still much better off than the GHC gitlab, they must be doing some clever blocking
2026-06-28 14:42:29 +0000 <tomsmeding> jaror: it's possible that forgejo is faster than gitlab, which would alleviate some of the pressure, but if it's just 2x as fast, that just postpones the problems until scraper volume doubles
2026-06-28 14:41:45 +0000fp(~Thunderbi@46-133-26-225.mobile.vf-ua.net) fp
2026-06-28 14:41:26 +0000fp(~Thunderbi@46-133-26-225.mobile.vf-ua.net) (Quit: fp)
2026-06-28 14:37:13 +0000 <jaror> For me it is ~10s per request but some other regularly have to wait minutes or even get errors.
2026-06-28 14:37:12 +0000 <tomsmeding> the problem with the GHC gitlab is that you genuinely use a bunch of the features of gitlab, including and not limited to the wiki
2026-06-28 14:36:42 +0000 <tomsmeding> but that holds for github too, perhaps even more, though github probably has micro$oft resources for blocking bad actors
2026-06-28 14:36:18 +0000 <tomsmeding> it _is_ true that my experience with gitlab so far, also outside GHC's instance, has been "it's kinda slow"
2026-06-28 14:35:43 +0000 <jaror> That might be the case, but in the end it is unacceptably slow. GitHub and Codeberg do seem to manage better.
2026-06-28 14:35:17 +0000 <tomsmeding> at some point, if there are enough requests, there's not much to do except blocking based on the sender of the request, which, with the botnets, is infeasible except for big centralised companies
2026-06-28 14:34:08 +0000 <tomsmeding> I wouldn't be surprised if the GHC gitlab is just spammed much more than git.tomsmeding.com, if only because there's more stuff on there to spam in the first place
2026-06-28 14:34:03 +0000chromoblob(~chromoblo@user/chromob1ot1c) chromoblob\0
2026-06-28 14:33:37 +0000 <tomsmeding> are we sure that the problem is that gitlab is slow?
2026-06-28 14:33:29 +0000 <tomsmeding> right
2026-06-28 14:33:19 +0000chromoblob(~chromoblo@user/chromob1ot1c) (Ping timeout: 264 seconds)
2026-06-28 14:33:14 +0000 <jaror> The admins (mainly mangoiv) is aware and has tried a lot, but nothing seems to work very well.
2026-06-28 14:32:10 +0000 <tomsmeding> and by whom
2026-06-28 14:32:02 +0000 <tomsmeding> what exactly is being spammed when the site is slow
2026-06-28 14:31:49 +0000 <jaror> Where "slow" means >10s response times
2026-06-28 14:31:43 +0000 <tomsmeding> I think to diagnose this you'd need info from the server admins about what's going on in the server logs
2026-06-28 14:30:20 +0000 <jaror> It seems Anubis might be causing some of the issues, but yeah even with anubis the GitLab is slow sometimes.
2026-06-28 14:28:46 +0000 <tomsmeding> (in the latter case, hand-rolling a portal and changing it once in a while might help, but not in the former case)
2026-06-28 14:28:16 +0000 <tomsmeding> and if so, do we have any idea if they just have a generic browser emulator or if they have special-cased code for anubis
2026-06-28 14:27:22 +0000 <fgarcia> i think something else people could try are hosting git and also using cloudflare or anubis
2026-06-28 14:27:21 +0000 <tomsmeding> jaror: are they getting through anubis?
2026-06-28 14:23:37 +0000 <tomsmeding> I don't know if they're spamming it, because I as a user of that VPS (also for other things) don't notice it :p
2026-06-28 14:22:48 +0000 <tomsmeding> Though maybe there just isn't enough on there!
2026-06-28 14:22:38 +0000 <tomsmeding> zero
2026-06-28 14:22:16 +0000gmg(~user@user/gehmehgeh) gehmehgeh
2026-06-28 14:22:06 +0000 <jaror> I'm mainly asking because the situation with the GHC GitLab is getting critical.
2026-06-28 14:21:27 +0000 <jaror> tomsmeding: have you taken special precautions against scrapers?
2026-06-28 14:21:06 +0000gmg(~user@user/gehmehgeh) (Ping timeout: 245 seconds)
2026-06-28 14:20:16 +0000 <jaror> mauke*
2026-06-28 14:19:43 +0000haritz(~hrtz@user/haritz) haritz
2026-06-28 14:19:43 +0000haritz(~hrtz@2a01:4b00:bc2e:7000:d5af:a266:ca31:5ef8) (Changing host)
2026-06-28 14:19:43 +0000haritz(~hrtz@2a01:4b00:bc2e:7000:d5af:a266:ca31:5ef8)
2026-06-28 14:18:41 +0000 <jaror> mauks: yes, I agree, but the 93% uptime makes it hard to convince others.
2026-06-28 14:17:36 +0000vanishingideal(~vanishing@user/vanishingideal) (Ping timeout: 246 seconds)
2026-06-28 14:17:24 +0000 <mauke> quite happy with codeberg so far
2026-06-28 14:17:02 +0000 <tomsmeding> jaror: my cgit instance is doing fine, although that of course is _just_ git, no issues, PRs, etc.
2026-06-28 14:17:00 +0000 <mauke> github is an ai scraper
2026-06-28 14:11:46 +0000 <jaror> I guess that's a new advantage of Darcs: the bots don't know how to clone your repo :)
2026-06-28 14:10:41 +0000 <jaror> It seems like github is the only place left that can provide some resistance to the ai scraper ddos