XSF Discussion - 2018-04-04


  1. Maranda

    nice and so I managed to send the error on the wrong stream...

  2. Maranda

    luckily it's fixed, all the blame goes on the flow scheme in {xep 220} that always confused me to hell.

  3. Bunneh

    Maranda: Server Dialback (Standards Track, Draft, 2015-03-12) See: https://xmpp.org/extensions/xep-0220.html

  4. SaltyBones

    Saw this on hn last night and had never heard of it: https://tox.chat/

  5. Ge0rG

    https://news.ycombinator.com/item?id=16750265 that comment made me really sad. And it doesn't even look like sarcasm...

  6. jonasw

    ow

  7. edhelas

    > Prosody is very simple to set up, ejabberd can scale to infinity.

  8. edhelas

    … and beyond

  9. Ge0rG

    Faster than the speed of light!

  10. jonasw

    objection!

  11. edhelas

    erlang makes your CPU runs faster

  12. Maranda

    That doesn't compute

  13. Maranda

    And I broke bidi ™️

  14. jonasw

    maybe you need a better compute-r

  15. Maranda

    How I don't have a clue (yet) ™️

  16. Maranda

    Aka never try to untagle messy code at 2:30 am.

  17. Maranda

    s/untagle/untangle/

  18. Maranda

    And found it

  19. dwd

    Maranda, I've got a broken bidi in Metre as well. It's deceptively tricky to get right.

  20. Maranda

    dwd: it's not bidi being broke, I broke mod_s2s which broke everything not just bidi, just because "mark connected" expects an event to return "true" and that is not there anymore yay \o/

  21. dwd

    Oh, fun.

  22. Maranda

    Let's hope no s2s stream closes before I'm home again.

  23. Maranda

    s/mark connected/make authenticated/

  24. Maranda

    dwd, good news is that I think I _finally_ managed to fix and untangle (most of) that mess that is s2s/dialback code.

  25. Maranda

    (Dialback expecially)

  26. dwd

    Metre is basically all S2S code, so I can appreciate that.

  27. Zash

    dwd: Have you tried Metre with Prosody ?

  28. Maranda

    And there we go...

  29. Ge0rG

    The more I think about the account-MUC-bouncer and some kind of extended notification mechanism, the more I like it.

  30. Maranda

    :O

  31. jonasw

    have you figured out a proper way of dealing with MUC service reboots and s2s errors yet?

  32. Ge0rG

    jonasw: there is none. But at least the server should know when s2s errors happen

  33. jonasw

    Ge0rG, syncing from one MAM to the other could work.

  34. Ge0rG

    MAM-to-MAM transfer!

  35. jonasw

    it doesn’t solve the JID hack tohugh

  36. jonasw

    DMA, direct MAM access.

  37. Ge0rG

    which JID hack?

  38. jonasw

    the JID hack which is called MUC

  39. Ge0rG

    Hm. Maybe.

  40. jonasw

    you know, where resources of a JID represent entirely different identities

  41. Ge0rG

    I've learned to live with that hack

  42. jonasw

    yeah, and you love the difficulites arising from PMs and such due to that :)

  43. jonasw

    the MSN things

  44. Zash

    Querying your contacts MAM for your own messages there could be ... a thing.

  45. Ge0rG

    Hm. Why doesn't my yaxim synchronize any more.

  46. Zash

    But you know, that's in "Skype behaviour I hate" land. Don't do it!

  47. Maranda

    Does it crash during startup?

  48. Ge0rG

    Maranda: no, it fails to join IRC channels.

  49. Maranda

    😎

  50. Zash

    Gotta crash a few times while querying your contacts online clients for history.

  51. Zash

    And get the history in random order.

  52. Maranda

    \o/

  53. Kev

    > But at least the server should know when s2s errors happen Nope, not really.

  54. Ge0rG

    Kev: *better than the client

  55. Maranda does a 🤔 at https://tools.ietf.org/html/rfc6120#section-8.3.3.4 and another 🤔 at https://tools.ietf.org/html/rfc6120#section-8.3.3.11

  56. Kev

    From the school of 'that's not saying much' :)

  57. Maranda

    I'm not sure I should use either when rejecting a db:request from a blacklisted entity, I suppose not-allowed would be more appropriate.

  58. Maranda guesses he should map all the rfc conditions rather than just those suggested in the xep.

  59. Maranda

    Dave Cridland, well bidi on dave.cridland.net doesn't look broken to me at a first glance at least fwiw

  60. Maranda

    (and/or dwd XD)

  61. Dave Cridland

    Maranda, It's not always broken. There's some sequence which causes it to choke.

  62. Maranda

    Dave Cridland, I think I can make an educated guess some of those involve DB 🤣

  63. Dave Cridland

    Maranda, To some extent. But I think it's more that it sometimes removes the session that's just gone bidi by accident. I thought I'd stopped it, but I think there's still some odd cases.

  64. Maranda

    and there I knew something was amiss, *hates being right*

  65. Maranda

    Dave Cridland, Does Metre do some sort of "caching" magic if I previously authenticated a stream via SASL external and I close the stream and reopen a new one and use dialback instead that time?

  66. Maranda is playing to try and trigger verification of keys with bidi but that doesn't seem to happen.

  67. Dave Cridland

    Maranda, Not sure I know what you mean. It doesn't change behaviour dependent on previous sessions (yet).

  68. Dave Cridland

    Maranda, But it'll respond to dialback based on whether it can authenticate the session. Not whether it can authenticate it via dialback.

  69. Maranda

    Dave Cridland, I don't see it verifying dialback keys that's what I mean.

  70. Dave Cridland

    Maranda, Right, but if the remote session has a valid certificate, there's no need.

  71. Dave Cridland

    Maranda, See XEP-0344.

  72. Maranda

    hmm ok noted

  73. Maranda

    since there's scarce bidi deployment out there I wanted to check if Metronome actually opened a new stream to verify keys (which I somehow doubt it doing on outgoing streams for some reason) :)

  74. Maranda wasn't aware of 344 either.