XMPP Council - 2018-04-18


  1. Dave

    Afternoon all.

  2. SamWhited

    o/

  3. Dave

    So, it's time.

  4. Dave

    1) Roll Call

  5. Dave

    SamWhited, daniel Kev?

  6. Kev

    Here.

  7. Dave

    (We seem to be a Ge0rG down)

  8. Dave

    Oh, and a daniel down as well?

  9. Dave

    Hmmm. Seems so. So we're at the minimum for quorum.

  10. Dave

    2) Last Minute Agenda Bashing

  11. Dave

    So Tedd Sterr beat me to an agenda this week, which meant that I didn't really check if there was anything else to add to it. Is there?

  12. SamWhited

    Not that I'm aware of.

  13. Dave

    3) XEP-0050 Execute Issue

  14. Ge0rG

    Whoops, sorry.

  15. Dave

    Kev, do you want to kick this one off?

  16. Kev

    Not desperately. I'm waiting for everyone else to have digested it properly and to say something.

  17. Kev

    I'm assuming folks haven't.

  18. Ge0rG

    Kev: everyone else is waiting for your expert opinion

  19. Kev

    Ge0rG: As of quite a few weeks ago now, everyone else was going to go away and try to understand it.

  20. Kev

    This is gnarly, I don't think me just suggesting text and people voting based on it looking reasonable is the right thing here.

  21. SamWhited

    I have read back through this one now, but I also don't remember what the original problem was.

  22. Kev

    SamWhited: That you can have an illegal state because execute is overloaded in weird ways.

  23. Dave

    As I understand it, it's that there is a silly-state in XEP-0050 whereby Execute effectively points to an action that doesn't exist.

  24. Kev

    You have an execute action, but you also have an execute attribute that sets a default action.

  25. Kev

    And that default action with the execute attribute isn't to use the execute action, which might be invalid.

  26. SamWhited

    ah yes, thanks. I read back over this, I did not read back over the two proposed solutions from last time, I should probably have done that too.

  27. Kev

    Dave: It's slightly more complicated than that, because of the weird overloading.

  28. Dave

    So specifically, §3.4 is the weird state, right?

  29. Kev

    https://github.com/xsf/xeps/pull/598 was how I was trying to address this.

  30. Dave

    That is, the default for the "execute", which is a sort of action-alias, is "complete" unless any actions are specified, in which case it's "next", which is not always present.

  31. SamWhited frantically reads this one and pretends he did it yesterday

  32. Kev

    Dave: Yes. Except that execute isn't really an action-alias, it's a default selected button on the form, and the execute action is just weird.

  33. Kev

    I tried to sort this out by tying the two together (see the first substantive line of the PR).

  34. Dave

    Kev, Now what i can't remember is why you weren't happy with your own PR.

  35. Kev

    Because of the way that execute and execute aren't currently the same thing.

  36. Kev

    And I want everyone to think carefully about whether my PR that tries to make execute and execute the same is going to break anything.

  37. Kev

    It wasn't that I believed my PR to be wrong, but that I think this one is gnarly and should have proper understand before folks vote.

  38. Kev

    It wasn't that I believed my PR to be wrong, but that I think this one is gnarly and should have proper understanding before folks vote.

  39. Dave

    Ah, I see.

  40. Kev

    Right now if you have execute='complete' and you run the execute command, what you run is 'next'.

  41. Kev

    And when you have no 'next' action that's clearly all sorts of messed up.

  42. Dave

    Wait, what?

  43. Dave

    If you have execute="complete", you run "complete", surely?

  44. Kev

    Hahaha, why would you do that?

  45. Dave

    Oh. No, it's not is it?

  46. Kev

    No, if you have execute='complete', and there is no 'next' action, and you run 'execute' what you are doing is running the non-existent command 'next'.

  47. SamWhited

    This was part of the confusion last time too; I'm inclined to say that whatever the final PR does it should do everything in its power to simplify the logic here.

  48. Kev

    SamWhited: I believe that's what my PR does. It would mean that where you have execute='complete', running <execute/> would be equivalent to <complete/>

  49. Dave

    Kev, OK. I now understand your PR in an entirely new light.

  50. Kev

    But this is a change to some pretty whacky implications and I want people to really understand what the issue is here before blithely accepting my suggestion based on believing that this bit of xep50 is less crazy than it really is :)

  51. Dave

    Kev, I also think it's potentially wrong. We should probably deprecate the execute action instaed.

  52. Kev

    Dave: I think that may well be the better solution.

  53. Dave

    So the case that concerns me is where execute="complete" but a next action exists; your proposal changes that I believe.

  54. Kev

    Exactly.

  55. peter

    OT: the other day I found my marked-up paper copy of XEP-0050 from ~8 years ago :-)

  56. Kev

    peter: Does it fix this? :)

  57. Dave

    OK, so yes, I think we should be deprecating the execute action on the basis that it doesn't do what I thought it did after several years.

  58. SamWhited

    That also seems like a good solution to me; especially since multiple people have been confused in a similar way here.

  59. Dave

    SamWhited, Indeed.

  60. SamWhited

    Unless peter's notes have an elegant solution written in the margin

  61. peter

    heh I'll take a look

  62. Kev

    Alas, the solution was too big to fit in the margin.

  63. Dave

    "There is a simple solution to this, but alas this margin is too small to contain it"

  64. Dave

    :-)

  65. Dave

    OK, let's move on.

  66. Dave

    Actually, given the time, do we want to discuss stream compression at all?

  67. Kev

    Newp.

  68. SamWhited

    I've got time today; but whatever you all want to do. I don't have much to say on the matter anyways.

  69. Dave

    Right, I think that's fine - we'll skip the remaining items. I'm not entirely sure what #577's status is anyway.

  70. Dave

    I'll commit to sorting that out for next week.

  71. Dave

    6) Outstanding Votes

  72. Dave

    I'll sort mine out after the meeting - Ge0rG, do you have a view on im-ng?

  73. Ge0rG

    Dave: sorry, not yet

  74. Dave

    Ge0rG, OK. You've a week before it goes through anyway.

  75. Ge0rG

    I'm currently way over my head in a long-range home relocation

  76. Dave

    7) AOB

  77. Dave

    Ge0rG, Exciting stuff.

  78. peter

    Alas, my notes on XEP-0050 seem to be mostly editorial, with a bit of content about multi-stage vs. single-stage command sessions, and caching.

  79. Dave

    (Assuming no AOB)

  80. Dave

    8) Next Meeting

  81. Dave

    Next week same time?

  82. Kev

    I'm almost certain not to make next week.

  83. Ge0rG

    +1W should work for me

  84. Dave

    SamWhited, still working for you?

  85. Dave

    (Assuming so)

  86. Dave

    9) Ite, Meeting Est.

  87. Kev

    Thanks all.

  88. Kev

    I'm glad I was finally able to share the horror on 50.

  89. Dave

    Thanks all. Sorry for my disorganisation, I'll get things back on track for next week.

  90. SamWhited

    Sorry, got distracted. Works for me. Thanks all!

  91. Dave

    Kev, Thanks for your patience here. I'd clearly been missing the key issue for ages now.

  92. flow

    > Kev> No, if you have execute='complete', and there is no 'next' action, and you run 'execute' what you are doing is running the non-existent command 'next'.

  93. flow

    Don't you run whatever execute points to in this case, e.g. 'complete'? Or am I missing something?

  94. SamWhited

    heh, that's what this whole discussion was about. You would think that, wouldn't you?

  95. flow

    Yeah, mostly from "The "execute" attribute defines which of the included actions is considered the equivalent to "execute" for this stage." in xep50

  96. flow

    It's been a while that I looked into it, but I gave a comparision of Kev's and my suggestion at https://github.com/xsf/xeps/pull/591#issuecomment-373313491 from my POV

  97. Dave

    flow, To my intense surprise, that's not what §3.4 actually says.

  98. Dave

    flow, It says that "3. If there is an <actions/> element [...] The action "execute" is always allowed, and is equivalent to the action "next".

  99. Dave

    flow, Hence my thinking is that the "execute" action is a disaster, and we should deprecate it (ie, SHOULD NOT use the action "execute").