XMPP Service Operators - 2023-12-03


  1. nuegia.net

    moparisthebest, that link is correct

  2. nuegia.net

    agh, I can reproduce it reliably.

  3. nuegia.net

    I recently upgraded to freebsd 14 and I will test if the bug is still reproducable there too

  4. nuegia.net

    during the required kernel update reboot I was able to reproduce the bug on 13.2 again as rc killed the jails

  5. maike

    I need to make a manifesto on weekly reboots

  6. maike

    who will sign it?

  7. nuegia.net

    why?

  8. nuegia.net

    maike, I won't sign it. I only intentionally reboot servers if there's a kernel update.

  9. nuegia.net

    what is your reasoning?

  10. Squeaky Latex Folf

    To recover from cosmic rays

  11. Squeaky Latex Folf

    To recover from cosmic rays?

  12. Squeaky Latex Folf

    Not all of us have the luxury of ECC memory

  13. nuegia.net

    I reccomend using ECC memory for servers or any computer system that does something important.

  14. moparisthebest

    maike: but then how would we have uptime competitions

  15. agh

    > agh, I can reproduce it reliably. Without iocage?

  16. Squeaky Latex Folf

    Or you can "save the planet" and recycle some of the many Sandy Bridge laptops out there as a server like I do

  17. Squeaky Latex Folf

    It's not professional by any means, but it's insanely cheap. Often enough it's not even the laptop that's the bottleneck

  18. Squeaky Latex Folf

    It should be more powerful than any Raspberry Pi

  19. Squeaky Latex Folf

    This practice is pretty common surprisingly too. ModFM is hosted on some old laptop as well

  20. Squeaky Latex Folf

    This practice is pretty common surprisingly too. ModFM from modules.pl is hosted on some old laptop as well

  21. nuegia.net

    ECC memory being rare, scarce, or expensive is a completely fabricated falsity created by Intel Corporation. Even the IBM PC had memory parity mode. Only modern-ish Intel consumer chipsets prevent you from posting if you have installed ECC memory completely artificially in an attempt to get you to buy the Xeon version, which is more or less the same exact cpu with all the features endabled.

  22. nuegia.net

    AMD's consumer platform allows you to use ECC memory and doesn't stop the computer from POSTING if it's detected.

  23. nuegia.net

    Even if the MMU didn't support that, you could just not use the extra chip on the ram stick. There's absolutely no reason to implement a check if it's ECC and stop posting on consumer products.

  24. nuegia.net

    boycot Intel CPUs.

  25. Squeaky Latex Folf

    AMD started reclaiming market share since they launched Ryzen in 2017. I still use hardware from 2012-2014 as my daily drivers, where Intel had a near monopoly

  26. moparisthebest

    Squeaky Latex Folf: even the rpi5 ?

  27. agh

    nuegia.net, Are you aware of https://www.youtube.com/@bhyvecon? If it is up your alley, you could join a jail(8)/zones(7) call, and you might be able to get started on some debugging.

  28. agh

    nuegia.net, those bhyvecon Videos are really awesome, and Michael Dexter is often the faciliator, and Dave Cottlehuber often makes an appreance, if just to make some witty remark.

  29. Squeaky Latex Folf

    » [02:27:12] <moparisthebest> Squeaky Latex Folf: even the rpi5 ? RPi 5 actually looks somewhat competitive, but I don't know how well it would compete to a 2012 x86 chip. Maybe there was a lot of IPC that ARM has gotten within the decade.

  30. agh

    There is some very talented and probably some of the most helpful people I have seen in the BSD community on those bhyvecon vids, often they are live debugging some issue someone has, there is lots of screen sharing, so you might be able to do some kernel dump debugging via that.

  31. maike

    moparisthebest, can't add 32GB ram to the rpi5 for $10

  32. moparisthebest

    The rpis use far less electricity

  33. moparisthebest

    For the same workloads

  34. theavidhorizon

    arm is great for this

  35. fjklp

    How would I go about checking for the /.well-known/host-meta of a xmpp server?

  36. fjklp

    well, looks like I figured it out. Turns out the first few servers I tried it on just didn't have the record, I guess.

  37. fjklp

    Is there any good excuse on having a short time limit on in-band account creation? I see this as a big common problem.

  38. fjklp

    It's as if we are trying to make people hat xmpp in the first few seconds of using it

  39. fjklp

    It's as if we are trying to make people hate xmpp in the first few seconds of using it

  40. fjklp

    What is the worst thing that could happen if the in-band account timeout were set to 1 hour?

  41. fjklp

    What is the worst thing that could happen if the in-band account setup timeout were set to 1 hour?

  42. fjklp

    A user should never be rushed, nor have their time wasted by having to redo things

  43. fjklp

    There should be some XSF recommendation for this

  44. fjklp

    I see nothing in https://xmpp.org/extensions/xep-0077.html

  45. sch

    Licaon_Kter (Fri 24 Nov 2023 10:45:18) AM > sch: and you do what with that JID? Talk where? Anywhere like a normal one? This is an example case where session (temporary) JIDs are useful https://github.com/thepug/Speeqe/wiki/SpeeqeInstances

  46. sch

    Licaon_Kter (Fri 24 Nov 2023 10:45:18) AM > sch: and you do what with that JID? Talk where? Anywhere like a normal one? This is an example case where session (temporary) JIDs are useful https://github.com/thepug/Speeqe/wiki/SpeeqeInstances (web based groupchat)

  47. sch

    https://metajack.im/2008/10/02/speeqe-is-open-to-all/

  48. Menel

    I don't see your full conversation, but many server use temp anon jids for a webchat

  49. Licaon_Kter

    sch: yes, then again it's not an anon account, but a "anon muc member" which imho it's a bit different

  50. sch

    As I recall, by looking into Jappix instances and servers running those instances, both were the same.

  51. Licaon_Kter

    fjklp: > Is there any good excuse on having a short time limit on in-band account creation? I see this as a big common problem. You mean that the user thinks about the password too much? What's the downside, redo registration but now it's faster since they down think of the pass that long?

  52. sch

    I'm not sure I've examined speeqe.com

  53. Licaon_Kter

    fjklp: > Is there any good excuse on having a short time limit on in-band account creation? I see this as a big common problem. You mean that the user thinks about the password too much? What's the downside, redo registration but now it's faster since they don't think of the pass that long?

  54. fjklp

    Licaon_Kter: could you reword that? I don't follow.

  55. Licaon_Kter

    fjklp: reword your issue :)

  56. fjklp

    are you joking?

  57. Licaon_Kter

    > Licaon_Kter: could you reword that? I don't follow. Are you?

  58. fjklp

    no

  59. maike

    the only timeout I've seen on IBR is for the captcha

  60. maike

    regular IBR is async and can't have a timeout afaik

  61. Licaon_Kter

    Then? What's the issue? How come it's too short? At which stage?

  62. fjklp

    I just tested psi+ and it does it the same as gajim

  63. maike

    timeout on solving the captcha makes some sense, eg. for the model of somone running tesseract, but not for a proper captcha solving service

  64. Licaon_Kter

    Does what?

  65. Licaon_Kter

    It's about capchas? You didn't say that fjklp

  66. Licaon_Kter

    In-band does not mean captchas, it can have them but not all servers do.

  67. Menel

    A complaint to the server admins where you encountered the problem would make sense maybe. I've only registered on servers with conversations some years ago where I didn't see any problem. But I can't even remember if capchas were involved

  68. ☭Mike Yellow

    A newly register user may ponder “what username and password should I set” when they get the register rable. Then they reach the time limit. When they notice it, they will prepare the username and password before getting a new table, if they want to continue registration.

  69. Licaon_Kter

    Not sure any info is even sent before pressing register...

  70. fjklp

    > A complaint to the server admins where you encountered the problem would make sense maybe. > I've only registered on servers with conversations some years ago where I didn't see any problem. But I can't even remember if capchas were involved It's ubiquitous and needs to be solved at the root, which is default settings on servers. The best thing I could think of to influence that is for an official recommendation, perhaps via XFS.

  71. Licaon_Kter

    fjklp: > It's about capchas timeouts? ????

  72. ☭Mike Yellow

    A newly register user may ponder “what username and password should I set” when they get the register table. Then they reach the time limit. When they notice it, they will prepare the username and password before getting a new table, if they want to continue registration.

  73. fjklp

    I'm not sure. Many have captchas. I don't know if I've experienced a timeout without a captcha. I never heard until now that the captcha is what makes the difference.

  74. Menel

    Prosody has a timeout for unauthenticated users, so one can't use up all connections the machine can have open. I think this would trigger on a not registered account too.

  75. fjklp

    It's not as if I get a message that says "Your captcha has timed out"

  76. Menel

    I think it defaults to 5min in prosody

  77. fjklp

    5 minutes is far better than what I usually see

  78. fjklp

    but I think we could go hitgher

  79. fjklp

    but I think we could go higher

  80. Menel

    I would suggest you write to the admins where you have a bad experience and tell them fjklp. As I said.. It's not a default what you see but an admin choice

  81. maike

    pick another server if you don't like it?

  82. fjklp

    It's ubiquitous and needs to be solved at the root, which is default settings on servers. The best thing I could think of to influence that is for an official recommendation, perhaps via XFS.

  83. fjklp

    what's the default on ejabberd?

  84. Menel

    I think it is already solved. Default server config doesn't have this. Also its less about the protocol per se. More a modern xmpp recommendation.

  85. Menel

    But for most the recommendations is invite only, and capchas are bad in any case

  86. fjklp

    well I'll be damned, it seems to work differently in Dino

  87. fjklp

    I don't get it, it seems like there is a longer timeout on loading the captcha in the browser when using dino

  88. fjklp

    ok, this is probably a gajim bug

  89. fjklp

    but I'm not sure yet

  90. sch

    > So i've made this: llama@decent.im . Enjoy! andrey.utkin, how did you implemented the Last Message Correction with your bot?

  91. sch

    > So i've made this: llama@decent.im . Enjoy! andrey.utkin, how did you implement Last Message Correction with your bot?

  92. sch

    And what does it take to establish an llama instance like you did?

  93. andrey.utkin

    Just launch llamafile

  94. andrey.utkin

    https://github.com/decent-im/llama-bot source

  95. sch

    I'm am new to llama. I might have read on it, and the first time I've ever used it was with your bot.

  96. andrey.utkin

    After one day of stupid stumbling, attempts to do it in lua...

  97. sch

    I've looked into the code, but I didn't fins wher and how you managed to implement the LMC

  98. sch

    I've looked into the code, but I didn't find where and how you managed to implement Last Message Correction.

  99. andrey.utkin

    sch, refreshingly simple because is the explicit feature and the goal of a specific slixmpp plugin and a XEP: https://github.com/decent-im/llama-bot/commit/b9d16bf18e69e0f058e2338599dbd80efdd1225c

  100. Squeaky Latex Folf

    nuegia.net: S2S from disroot.org to nuegia.net fails with: There was an error getting items for nuegia.net. Reason: Remote server timeout. A remote server or service specified as part or all of the JID of the intended recipient (or required to fulfill a request) could not be contacted within a reasonable amount of time. Server-to-server connection failed: unsupported-stanza-type