XSF Discussion - 2023-03-06


  1. Menel

    https://xmpp.org/extensions/xep-0408.html seems to be on last call that ended 2021 according to the red text. I suppose some new status belongs there

  2. Menel

    https://xmpp.org/extensions/xep-0401.html seems to be on last call that ended 2021 according to the red text. I suppose some new status belongs there

  3. Daniel

    There is no way to map an occupant id to the members, admins list is there?

  4. Zash

    On its own, no

  5. Zash

    You'd have to observe <presence> or whatever

  6. MattJ

    I wondered recently about just inserting it into the affiliation responses

  7. Zash

    Outch, the O(n^not fun)

  8. MattJ

    I think we cache it thou... oh right, not for offline folk

  9. Zash

    affiliations ≠ occupants

  10. Daniel

    > You'd have to observe <presence> or whatever Yes I was thinking about offline contacts

  11. Link Mauve

    Also in XEP-0463.

  12. Zash

    Hm, but they're based on bare jid?

  13. Zash

    Brain too full of OAuth to swap in another XEP

  14. Daniel

    Yes they should be based on bare jid

  15. Zash

    Then it should be doable (and relatively cheap) to include anywhere we have affiliation

  16. Daniel

    if we want occupantId to be the one stable thing that identifies people in muc it should probably be everywhere

  17. Daniel

    so I can map the different sources of "people who are in a muc" and merge them properly

  18. Daniel

    and look up the "real sender"

  19. Zash

    (jid, nick, occupant-id) tripplets everywhere

  20. Ge0rG

    (jid, nick, occupant-id, origin-id, stanza-id)

  21. Daniel

    I’m not the one who killed mix

  22. Zash

    Ge0rG, fastening-context, version number?

  23. Zash

    Daniel, MIX relates to this how?

  24. Ge0rG

    MIX is dead?

  25. Zash

    Weren't we all going to implement MIX over the weekend?

  26. Daniel

    mix has stable participant ids by default

  27. Daniel

    no resource vs occupant id

  28. Zash

    I thought it had real JIDs for everyone by default, and some proxy service

  29. Zash

    I thought it had real JIDs for everyone by default, and some proxy service (not included)

  30. Ge0rG

    real JIDs are the most stable IDs

  31. Daniel

    yes real jids are included too. but messages originate from room@service/participant-id

  32. Daniel

    and this maps to the item id of the participant node if you want to look up more information

  33. Zash

    Invent some MUC join options, one of which would be to end messages from room@muc.service/occupant-id

  34. Zash

    Invent some MUC join options, one of which would be to send messages from room@muc.service/occupant-id

  35. Zash

    Rewrite the nick into XEP-0172 elements or somesuch

  36. Daniel

    > Invent some MUC join options, one of which would be to send messages from room@muc.service/occupant-id I could live with that

  37. Zash

    And then more and more MUC options until finally we have a "just use MIX" MUC option :)

  38. Daniel

    When I previously brought up something similar people argued that some clients might not support muc initiated nick rewrites

  39. Zash

    "Don't send presence" being another such option

  40. Zash

    Yeah, if we just do it without opt-in

  41. Zash

    Feature flags sure ... moves some pain around

  42. Daniel

    I think it would have to be a channel wide setting. Otherwise you can't rely on that as the one who is parsing the messages

  43. Zash

    Isn't that just "use MIX instead" then?

  44. Zash

    Wait nick rewrites? Isn't that a different thing

  45. Daniel

    Muc can tell you 'your resource (nick) is x now' instead of the one you tried to join with

  46. Daniel

    There could be a muc setting where the muc tells everyone your nick is occupant id now

  47. Zash

    I mean <presence><{muc}x><{fancy-new-options}swap-nicknames-and-occupant-ids/><//> (check disco first) and then you get a different kind of broadcasts

  48. Daniel

    Ah

  49. Daniel

    Mhh maybe

  50. Daniel

    That sounds complicated

  51. Daniel

    Because it would have to change the resources in the MAM response too

  52. Zash

    Sure, yeah.

  53. Daniel

    If people prefer something like that to mix I'm fine with it

  54. Zash

    I imagine it would be the same class of complexity as implementing MIX protocol on top of existing MUC code

  55. Peter Waher

    Hello all. (Sorry for cross posting…) I’m experimenting a bit with bridging OpenAI chat and the XMPP network, to see what this can be used for. If anyone is interested to try, you can chat with xmpp:chatgpt@lab.tagroot.io. Image generation can be tested by chatting with xmpp:dalle256@lab.tagroot.io, xmpp:dalle512@lab.tagroot.io or xmpp:dalle1024@lab.tagroot.io. For more info, or QR-codes to the above JIDs, https://lab.tagroot.io/Community/Post/OpenAI_integration_with_the_Neuron

  56. Peter Waher

    Hello all. (Sorry for cross posting…) I’m experimenting a bit with bridging OpenAI chat and the XMPP network, to see what this can be used for. If anyone is interested to try, you can chat with xmpp:chatgpt@lab.tagroot.io. Image generation can be tested by chatting with xmpp:dalle256@lab.tagroot.io, xmpp:dalle512@lab.tagroot.io or xmpp:dalle1024@lab.tagroot.io. For more info, or QR-codes to the above JIDs, https://lab.tagroot.io/Community/Post/OpenAI_integration_with_the_Neuron

  57. Peter Waher

    While the gateways above have minimal instruction, it is possible to fine-tune a model with up to 1 GB data using a dedicated instance it seems. It would be possible to make a bot, for instance, that can learn about all XEPs and then help people to find relevant XMPP-related information.

  58. Menel

    a fun endevaor. did you also try to put it in a muc?

  59. Peter Waher

    I was thinking about it, but didn’t get a clear idea on how it would operate. It would be annoying if it inserted itself into everything. Perhaps it could stay silent, and only monitor communication (to learn) and inject a comment/response if asked directly (or via a private channel)

  60. Menel

    react only if mentioned first of course

  61. moparisthebest

    I have a markov chains XMPP bot, millions of dev hours in AI development, no use case that isn't also filled by markov chains from 1906, they have played us for absolute fools

  62. pep.

    Yet another fad that can get you money from VCs I guess

  63. Menel

    we could let the two bots talk to each other 😛️

  64. moparisthebest

    https://code.moparisthebest.com/moparisthebest/ash is the source, you can play with it in a few MUCs, one is xmpp:conversations-offtopic-reloaded@conference.trashserver.net?join

  65. mentos124

    open source apps are the best

  66. Fishbowler

    I've written some open source apps that can disprove that :)

  67. moparisthebest

    Personal opinion: open source is required for an app to be good, but is not alone sufficient to be good, in fact nearly all software sucks :)