XMPP Service Operators - 2025-06-21


  1. tom

    Why does s2s use separate connections for outgoing and incoming? Is there a reason bidi is a community module and not part of base in prosody? Why would I not use bidi?

  2. tom

    What is the benefit to not using bidi?

  3. Brian

    It's not a community module, but it is a module, like just about everything else in Prosody.

  4. Brian

    Never mind... just spotted the community module for it. There's a base module called s2s_bidi.

  5. Menel

    It's not a community module anymore since some years. What you see is a remnant

  6. Menel

    I use it since over 5 years stable

  7. Menel

    Had not any ejabberd connection issues since some month now.

  8. jjj333_p [pain.agency]

    i dont think i have any mod_bidi enabled on mine, is there any particular advantage to it?

  9. jjj333_p [pain.agency]

    in particular if it comes with less overall bandwith to any measurable degree that would be nice

  10. Menel

    It's half the encrypted tcp connections, so a bit less memory. Some depending on the server size some MB saved.... I didn't measure it.

  11. Menel

    It's half the encrypted tcp connections, so a bit less memory. So depending on the server size some MB saved.... I didn't measure it.

  12. jjj333_p [pain.agency]

    ah. system resources isnt really the concern since i havent downsized since hosting matrix, however it would be nice to not have bandwith overages every month

  13. Menel

    I would think bandwidth comes from media upload / proxy, not the xmpp stream, additionally I don't think bidi saves anything relevant there.

  14. jjj333_p [pain.agency]

    a lot of it is ironically 404 http requests from matrix since it kinda crapped on me instead of getting to gracefully shut it down

  15. jjj333_p [pain.agency]

    media is prob a lot though

  16. jjj333_p [pain.agency]

    i checked and caddy is consistently the highest bandwith despite it just being that and a http file upload that only i upload to

  17. Menel

    What I thought

  18. Menel

    To save bandwidth you could drop the matrix traffic at the firewall Leven per IP maybe.... 😀

  19. Menel

    To save bandwidth you could drop the matrix traffic at the firewall level per IP maybe.... 😀 But I would think its not that much to send 404s

  20. tom

    pain

  21. Kris

    Best to drop the domain. Matrix servers will never stop trying

  22. nigel

    They never... that is a stupid design.

  23. Kris

    Indeed 🤷

  24. jjj333_p [pain.agency]

    tbh if i dont revive what exists, i might at least try to rejoin rooms so that i can evacuate them, that should at least cut down on the reason to federate

  25. jjj333_p [pain.agency]

    after that its just pointless federation and/or state resets

  26. Menel

    To be fair xmpp also retries as long as someone has a user in the server in their roster

  27. nigel

    Ah.

  28. Menel

    There is a workaround module for prosody to automatically unsubscribe users for non existing domains, but the admin has to do stuff for that and be aware

  29. jjj333_p [pain.agency]

    i would expect a quick connection attempt for a 1:1 dm to be a little more lightweight than mistakenly trying to contact my server for every event and presence update in any room i mightve been joined to

  30. jjj333_p [pain.agency]

    also tcp is a little easier to just block if need-be

  31. Kris

    > There is a workaround module for prosody to automatically unsubscribe users for non existing domains, but the admin has to do stuff for that and be aware :thinking-face: