XMPP Service Operators - 2026-05-04


  1. icebound.dev

    > hosting in ~the cloud~ somebody else's server doesn't eliminate the need for backups this is true, but it provides more resilience

  2. icebound.dev

    > reduncancy: the I in RAID stands for Inexpensive yeah apart from RAID 1 being the least efficient space wise... so the most costly...

  3. icebound.dev

    > backups expensive? how much is a thumb drive? flash drives are not reliable storage media, so you basically just made yourself look like an idiot :/

  4. icebound.dev

    > ddos immunity? bullshit Its a lot harder to flood a massive datacentre than it is to flood someones 100mbps home internet

  5. icebound.dev

    > flash drives are not reliable storage media, so you basically just made yourself look like an idiot :/ also to do 3 2 1 yes its pricy. 3 copies, 2 on site, 1 off site. Your main copy is usually for performance, so RAID 0 isn't uncommon for this. The backup is often more rigid, so RAID 1, 5, 6 or 10 is common. Your offsite copy, unless you can place two backup servers in two different places (one onsite, one off site), then you will likely use cloud, which can be costly. So yes proper backups cost a lot of money.

  6. icebound.dev

    thankfully this isn't much of an issue with XMPP, as most servers do not store much data, MAM and http file uploads usually expire and they take up the most storage.

  7. icebound.dev

    in any case this conversation should end, its not so much offtopic, but it isn't going anywhere useful.

  8. jjj333_p [pain.agency]

    > ur cooked i mean prosody (or probably any xmpp server impl) would run fine on like a semi-modern raspberry pi, or like a really old laptop, you only would really need to look out for storage speed, in particular for the http upload

  9. jjj333_p [pain.agency]

    and in theory theres no requirement that the http file upload is on the same device, and you could probably figure out some s3 setup

  10. jjj333_p [pain.agency]

    > Yeah I never send anyone at the public provider lists for this exact reason i usually try to recommend specific providers, however from my experience helping others (i havent used it directly) cheogram tends to just shove people on to chatterboxtown.us which is one of the worst providers?

  11. singpolyma

    it's been jabber.fr for awhile

  12. jjj333_p [pain.agency]

    both of which ive had to ban from my mucs from just tons of spam

  13. admin_cbt

    I've been working on keeping my server off the list https://xmppbl.org/stats

  14. hueso

    >> flash drives are not reliable storage media, so you basically just made yourself look like an idiot :/ > also to do 3 2 1 yes its pricy. > > 3 copies, 2 on site, 1 off site. it's actually 3 copies, on 2 different media. Flash drives are fine for that, no media is 100% reliable. RAID ain't no backup.

  15. TheCoffeMaker

    Dude ...most reliable media are DVDs when stored correctly

  16. hueso

    my old DVDs were eaten by bacteria

  17. hueso

    or fungi idk but there are missing pieces

  18. Nigel

    > Dude ...most reliable media are DVDs when stored correctly you mean tape

  19. Lilith

    > both of which ive had to ban from my mucs from just tons of spam jjj333_p [pain.agency]: I've been messing a bit with https://modules.prosody.im/mod_muc_auto_role.html . It seems like a promising solution

  20. icebound.dev

    > and in theory theres no requirement that the http file upload is on the same device, and you could probably figure out some s3 setup Http upload was designed so that it can be integrated into CDN's.

  21. icebound.dev

    > >> flash drives are not reliable storage media, so you basically just made yourself look like an idiot :/ > > also to do 3 2 1 yes its pricy. > > > > 3 copies, 2 on site, 1 off site. > it's actually 3 copies, on 2 different media. Flash drives are fine for that, no media is 100% reliable. RAID ain't no backup. 2 different media, usually means different servers, but yes, both copies are on site, I am not wrong. Also I know RAID is not a backup, but it provides further rigidity to your backups, and also higher performance for your production servers, especially if you are using spinning rust.

  22. icebound.dev

    which right now, unless you are really rich, NAND is not an option.

  23. icebound.dev

    in any case, flash drives are probably the worst storage medium after SD card.

  24. icebound.dev

    I have tried using the 07f.de contact a few days ago, I have heard nothing back from them.

  25. icebound.dev

    bus factor of 1.

  26. stratself

    does providers. provide any contact info about the instances' admins?

  27. icebound.dev

    > does providers. provide any contact info about the instances' admins? no.

  28. icebound.dev

    you mainly get it off their site

  29. icebound.dev

    when I get a chance I will implement whois for authbot, but getting jonas to review it might be difficult because I believe they are busy.

  30. Menel

    There are bots and clients that already do that. So if someone is curious about contact info, one doesn't have to wait for a specific bot.

  31. icebound.dev

    surre.

  32. icebound.dev

    sure.

  33. icebound.dev

    I am just saying it would be useful to be able to whois people in the channel, bidirectional. So right now for example I cant remember what server you host Menel, so I could whois you to see what server you were granted voice for, therefore represent. And then if say I was having issues with your server, and I didnt know who you were in this channel, I could whois and get the nick of the user. I think it could be really useful :D

  34. icebound.dev

    Sure I could get the admins from the server info manually, but it could be useful instead of asking "who hosts x server" in this channel, or "if the admin for x server in this channel"

  35. stratself

    just code a server module that responses to `!admin debug fetch-support icebound.dev`

  36. stratself

    just code a server module that respond to `!admin debug fetch-support icebound.dev`

  37. icebound.dev

    > just code a server module that respond to `!admin debug fetch-support icebound.dev` Thats less useful, because its omnidirectional. I can't then go search which admin in this channel represents what server. The idea came when people complained about asking people to stick their server into their nick :)

  38. icebound.dev

    in general its better to report issues publicly than PM pepole about issues.

  39. icebound.dev

    so for example if the admin of 07f.de was here, I could report it to them in public so other operators know if people are reporting issues with 07f.de that theres an ongoing discussion about it. Isn't that exactly the benefit of having an operators channel?

  40. stratself

    just make this channel non anonymous, problem solved

  41. icebound.dev

    that is true, but what about those who are not admins who spectate

  42. singpolyma

    there's not supposed to be any non operators here

  43. singpolyma

    but I agree the definition operator is a bit strained

  44. luca

    How would https://logs.xmpp.org/operators/2026-05-04 react to non-anonymous jids?

  45. luca

    (or how _should_ it react)

  46. icebound.dev

    > there's not supposed to be any non operators here I think it is useful for non-operators to spectate tbf

  47. icebound.dev

    the logs show it anyways, users should ideally be allowed to see the messages and keep track of issues

  48. icebound.dev

    XSF discussions doesnt require you to be in the XSF to join...

  49. icebound.dev

    in all honestly I do feel its stupid that non-operators cant report issues here, it means they contact their server operator to then report an issue with another operator...

  50. Menel

    Well, people like to Spam and report and talk about stuff that's not relevant for everyone. That's the reason. If it only affects one service then it should be talked to them for example

  51. Menel

    People have other things to do and don't want to read so much that doesn't affect them. This is already quite a bunch of traffic the last days for example. At one point people just don't read or leave

  52. icebound.dev

    > Well, people like to Spam and report and talk about stuff that's not relevant for everyone. That's the reason. > If it only affects one service then it should be talked to them for example thats not ideal. If I discover a bug with another server, and I report it to them directly, then another operator then discovers the same bug, and so on, the operator gets many reports. Meanwhile if someone publicly finds it here, other operators will see potential bugs their server might encounter, and it all also stop the admins being hit with multiple reports.

  53. icebound.dev

    > People have other things to do and don't want to read so much that doesn't affect them. This is already quite a bunch of traffic the last days for example. > At one point people just don't read or leave You dont get a notification do you, if I had an issue with your server I would mention you...

  54. Menel

    You got I the other way around. If there is an issue with my sever I expect a PM. Here it is for issues that affact many or all. Then nobody will mention individuals for such announcements

  55. icebound.dev

    > You got I the other way around. > If there is an issue with my sever I expect a PM. > Here it is for issues that affact many or all. Then nobody will mention individuals for such announcements welp I guess we differ on how we find the channel useful

  56. icebound.dev

    whats a point of an operators channel if its just for support with general setup?

  57. icebound.dev

    theres nothing worse than needing to track numerous PMs to report issues, meanwhile I can just ping the respective people here, and call it a day, much easier IMO.

  58. jjj333_p [pain.agency]

    > jjj333_p [pain.agency]: I've been messing a bit with https://modules.prosody.im/mod_muc_auto_role.html . It seems like a promising solution ooo this is the kinda thing ive been looking for, tysm Lilith [queer-spark.org],

  59. jjj333_p [pain.agency]

    > Http upload was designed so that it can be integrated into CDN's. yeah, but prosody doesnt seem to have native s3 support, i remember when i asked they said id probably have to do it myself. i wanna make my own external http upload component sometime though

  60. moparisthebest

    yes we need a 10,000th http upload server component :P

  61. jjj333_p [pain.agency]

    its somewhere on my long list of todos, may or may not ever happen

  62. Lilith [queer-spark.org]

    > ooo this is the kinda thing ive been looking for, tysm Lilith [queer-spark.org], No problem. Though, it applies to all group chats on a component, not just ones in moderated mode

    👍 1
  63. jjj333_p [pain.agency]

    i suppose one could have multiple muc components

  64. Lilith [queer-spark.org]

    I want to get https://modules.prosody.im/mod_muc_members_json.html or something similar working first. Otherwise, it's just going to create a lot of work for the mods on QS

  65. jjj333_p [pain.agency]

    > jjj333_p [pain.agency]: I've been messing a bit with https://modules.prosody.im/mod_muc_auto_role.html . It seems like a promising solution wait i just realized lol

  66. jjj333_p [pain.agency]

    https://downloadable.pain.agency/file_share/019df4e7-c7f3-7402-8ed1-552cfabdbf5d/2320e270-14a0-499d-a47d-ec16726590a4.png

    😁 1
  67. jjj333_p [pain.agency]

    irrelevant inconsistency, but funny

  68. icebound.dev

    > yeah, but prosody doesnt seem to have native s3 support, i remember when i asked they said id probably have to do it myself. i wanna make my own external http upload component sometime though jjj333_p [pain.agency], implement it yourself if you want it

  69. icebound.dev

    the state of S3 sucks right now

  70. icebound.dev

    Minio dropped its community edition

  71. icebound.dev

    and minio was basically the standard for S3 self hosting

  72. singpolyma

    There's a half dozen others though, including a maintained fork of minio

  73. icebound.dev

    > There's a half dozen others though, including a maintained fork of minio theres a maintained fork?

  74. icebound.dev

    please link, a non-profit I volunteer for is using minio, and they are currently vulnerable

  75. icebound.dev

    minio from Dec (final release) has numerous vulns in it

  76. singpolyma

    https://github.com/pgsty/minio I think this one

  77. icebound.dev

    alright cool

  78. icebound.dev

    it fixes a few of the vulns

  79. icebound.dev

    I will relay, thanks a ton singpolyma

  80. Kris

    Garage is a good replacement

  81. icebound.dev

    thx will check it out