Groupchat 3.0 development - 2026-03-05


  1. rdobovic

    Hi everyone! Is there some kind of draft specification for GC3, besides presentation from the XMPP summit?

  2. MattJ

    Not yet

  3. MattJ

    There are some things written, but it's spread out and incomplete

  4. MattJ

    GC3 is mostly at the idea/prototype stage at the moment, a specification will come

  5. rdobovic

    Oki thank you

  6. rdobovic

    Do you have a rough idea how much time will take for that to happen? Just curious.

  7. singpolyma

    Depends on many things. Is there a reason you're interested?

  8. rdobovic

    Nothing important. I recently started getting into XMPP and reading and reading the specs. I find the whole idea quite interesting, however comming from Discord, I am intrested to see if it will be possible to implement someting like Discord's role permission system with XMPP.

  9. rdobovic

    Nothing important. I recently started getting into XMPP and reading the specs. I find the whole idea quite interesting, however comming from Discord, I am intrested to see if it will be possible to implement someting like Discord's role permission system with XMPP.

  10. rdobovic

    Nothing important. I recently started getting into XMPP and reading the specs. I find the whole idea quite interesting, however coming from Discord, I am interested to see if it will be possible to implement something like Discord's role permission system with XMPP.

  11. moparisthebest

    XMPP probably already has what you are after but if not yes it's possible to add

  12. Kev

    Was it here I shared my protoprotoXEP for Discord-style hats-based ACLs the other day?

  13. rdobovic

    Don't know about before, but you send it to XSF Discussion this Tuesday when I asked a few questions there.

  14. rdobovic

    Don't know about before, but you sent it to XSF Discussion this Tuesday when I asked a few questions there.

  15. Kev

    You’ve seen it, at least, then :)

  16. rdobovic

    Yep, I read it, I'm looking forward to see it complete :)

  17. moparisthebest

    don't wait, implement it

  18. Kev

    The spec isn’t complete yet :)

  19. Kev

    Although I think completing it is probably mechanical at this point.

  20. moparisthebest

    no spec is ever complete without an implementation

  21. Kev

    There are degrees of incompleteness :)

  22. rdobovic

    Do you think it would be possible to make it compatible with current MUC specification?

  23. singpolyma

    Yes

  24. Kev

    I think it would be possible to use it in an implementation that also does MUC. MUC has its own (simple) ACL system, so it depends exactly what you mean.

  25. singpolyma

    MUC already has roles and hats

  26. Kev

    The aim was for it to be a clean migration from MUC ACLs to this system.

  27. rdobovic

    If I understood correctly, current MUC ties permissions to it's own roles and affiliations. So, to connect it with this new permissions system, I guess you could map MUC affiliations to required hats. This way normal operations with affiliations would work. But what would you do about roles?

  28. singpolyma

    Roles are basically particular permissions a participant has at the moment. So a hat could confer a role or not but this is up to implementation

  29. rdobovic

    So, would it be a problem if implementation ignored roles and assigned for example participant to everyone?

  30. singpolyma

    The MUC you mean? No that's definitely allowed

  31. rdobovic

    In that case, could implementation also ignore requests to change role of a user? So that all users always have a role of participant?

  32. moparisthebest

    yep

  33. rdobovic

    Great, just one more thing. Since implementation is free to grant any privilege set to each role, how current client implementations know if the client is allowed to, for example, send messages?

  34. rdobovic

    Btw, if I'm bothering you or diverging from the room topic too much, just say, I'll stop :)

  35. Menel

    Generally privileges are bundled with a role, as specified in the xep. I would guess current open clients would think they have these privileges and will get an error from the muc if they try something they can't do

  36. singpolyma

    A role is not the same as an affiliation

  37. singpolyma

    for example "visitor" role means "can read but can't send" it's maybe best to not think of it as a role but just as "how the server tells you if u can write"

  38. rdobovic

    Well, in that case I guess that, in the context of these hat based permissions, it would be the best to map user's role to the first role that covers all of user's permissions (granted by hats).

  39. rdobovic

    Anyway, thanks for all the answers!