XSF Discussion - 2019-06-12


  1. jonas’

    lovetox, I cannot join jdev@ because jabber.org has DH keys which are too small to be accepted by my server

  2. jonas’

    white is 255,255,255, so inverting gives 0

  3. pep.

    Dumb question about the whitespace thread on standard@, I've read this: "it's more efficient for battery usage to use push notifications", bit Google also has to maintain a TCP connection to push these notifications right? So they are eventually running into the same issues as us? How is that more efficient

  4. pep.

    Dumb question about the whitespace thread on standard@, I've read this: "it's more efficient for battery usage to use push notifications", but Google also has to maintain a TCP connection to push these notifications right? So they are eventually running into the same issues as us? How is that more efficient

  5. Holger

    pep.: I think this story of push notifications saving traffic originated from them being used as an alternative to crappy (HTTP) polling. The remaining point I see is that with push notifications, the OS can freely kill the app on resource shortage.

  6. Holger

    Plus maybe a little traffic optimization after all, if you take into account that TCP keepalives or thelike (I think theirs weird stuff done on mobile) are per connection.

  7. Holger

    But in general you're right, I think.

  8. Kev

    I would /assume/ (very possibly wrongly) that they're smart enough to not go spinning the radio all the way up for push notifs.

  9. Ge0rG

    Kev: I'm not sure this is how baseband modules work.

  10. Ge0rG

    So yes, it's a matter of keeping a single TCP connection vs one per app or even long polling / periodic polling by each app

  11. Ge0rG

    Also google has done significant work on making that connection stable, even under adverse conditions like very short NAT timeouts

  12. Ge0rG

    (they are measuring them and maintaining per network keep alive intervals)

  13. Kev

    Ge0rG: Well, whether you can control the baseband or not (Apple presumably can, at least), you can make your push notifs small enough to avoid significant wakeups, no?

  14. Kev

    But I'm working entirely on intuition here, no clue what I'm talking about.

  15. Ge0rG

    Kev: well, yes

  16. Zash

    Being smart about scheduling could help, in theory

  17. Ge0rG

    I don't think there are significant delays in delivering push to Android. Except for the trash calendar app on my Samsung phone, where all notifications come a day late

  18. MattJ

    iirc Samsung have their own push thing, or used to

  19. pep.

    So you have to go through Google and then Google goes through Samsung? :p

  20. pep.

    Or you skip Google and have to support two APIs ?

  21. Zash

    Haha

  22. Ge0rG

    I have no idea why it doesn't work properly. It was perfectly on time on LineageOS on my previous phone

  23. flow

    pep., batching the push notifications and takign the devices sleep cycle into account is the only minor argument I am able to come up with, but besides that, an XMPP connection could be as efficient as an google push connection

  24. flow

    but google does obviously neither for high priority FCM pushes, only probably for normal priority messages

  25. rion

    why jingle has only global tie-break error but not per content? What if we have multiple <content> elements in a request and only one of them is <tie-break>. Basically the same for <reason> which may be different

  26. flow

    rion, potentially no particular reason and just something the authors did not consider. From how I understand jingle, having per content tie-breaks sounds sensible

  27. rion

    I'm implementing transport-replace in Psi right now. It's mostly finished. But these problems with tie-break and necessity to send transport offer twice make me frustrated.

  28. pep.

    Can there be fulljids in the roster? (that is, a fulljid I'd be subscribed to)

  29. MattJ

    pep., iirc yes, you can (in 6121, possibly not 3921), but don't expect it to work how you want, you can't subscribe to a full JID

  30. Zash

    Are you sure? I thought it was bare JIDs only

  31. MattJ

    Zash, I can find no such restriction

  32. Link Mauve

    I also remember a discussion on that topic, but none of the conclusions.

  33. dwd

    To go back to this morning, the argument for push notifications over a TCP session comes down to that if all the pushes go through the same channel, this can be optimized in terms of combining radio activity. But I think it's a sketchy argument, given you'd have to introduce latency and things in order to combine them anyway.

  34. rion

    guys, what do you use to edit XEPs? To preview them quickly

  35. flow

    rion: sam build a inotify based make target that would build the html every time the xml changed

  36. flow

    I use something similar like that

  37. jonas’

    rion, I use a texteditor and my typical arrow-up-enter workflow

  38. jonas’

    with `make build/xep-1234.html`

  39. Zash

    `:make`

  40. rion

    ok. got it. thanks.

  41. lovetox

    lol jabber.ru sends russian text on the IBR form

  42. lovetox

    no idea what they want me to put into the field

  43. jonas’

    lovetox, I’m not in jdev@ because I can’t join due to Server-to-server connection failed: Connecting failed: sslhandshake error: dh key too small

  44. jonas’

    lovetox, I’m not in jdev@ because I can’t join due to Server-to-server connection failed: Connecting failed: ssl handshake error: dh key too small

  45. lovetox

    yeah i read your message

  46. jonas’

    ah ok

  47. jonas’

    you’ll have to live with that until jabber.org is fixed

  48. rion

    lovetox: try to type some russian =)

  49. lovetox

    thanks for the hint with background, i though for some reason 0,0,0 is white :D

  50. jonas’

    :)

  51. jonas’

    yeah, the jclib implementation is historically grown and a bit of a mess

  52. lovetox

    works good now for me, gajim has now also default avatar generation

  53. jonas’

    \o/

  54. jonas’

    awesome

  55. rion

    Пользователь = username Электронная почта = email Телефон = phone

  56. jonas’

    I need to get the text on semianon vs. nonanon in 392 and then it can be advanced IMO

  57. lovetox

    thanks rion i found my old account just now and rememberd the password

  58. jonas’

    (specifically, in a semianon room, you SHOULD use the nickname, even if you have privileges and can see the real JID; in a nonanon room you SHOULD use the real JID)

  59. lovetox

    yeah although this is bothersome to implement

  60. rion

    lovetox: so russian FSB won't follow you since phone number wasn't required previously

  61. lovetox

    my contact object does not have the information about the room settings the contact is in

  62. lovetox

    :D

  63. lovetox

    rion, yeah because anyone puts in the real number there :D

  64. jonas’

    lovetox, yeah, transferring that information will be a tad annoying

  65. jonas’

    but it might be good to do that anyways, considering that you don’t want to accidentally leak a real JID you only know because you are a privileged member of a room

  66. jonas’

    so carrying that information around might be good to present that bit ("you only see this because you are a moderator, tread carefully") in the UI

  67. Ge0rG

    lovetox: speaking of old accounts. There was a bunch of spam bots on Gajim.org recently. Do you have IBR enabled? Is it even meant to be a public server?

  68. rion

    > Accounts not used for more than 1 month are destroyed. > To register an account, go to this URL : <a href="https://gajim.org:5281/register_web" style="color:#000080">https://gajim.org:5281/register_web</a>

  69. lovetox

    please just send them to me, i will forward it to asterix to delete them

  70. rion

    mm now I have an account on gajim.org

  71. lovetox

    and yeah register on gajim.org is enabled, but i will question if this is necessary

  72. lovetox

    as the server is not maintained for users really

  73. Ge0rG

    So maybe somebody broke the captcha? But it was only two bots, that's not the typical pattern. They come in hundreds normally

  74. lovetox

    there is no captcha, just a "what is 7 + 9"

  75. lovetox

    probably very hard for bots

  76. Ge0rG

    I don't have the JIDs here right now, you just weren't online for a long time

  77. lovetox

    yeah i dared to go on holiday 2 weeks

  78. Ge0rG

    lovetox: you need a mobile xmpp client! 😜

  79. lovetox

    hm im still evaluating what the real price of having a smartphone is, could take a while

  80. jonas’

    I’m interested in your calculations

  81. rion

    $ expr $(wget -qO- https://gajim.org:5281/register_web | grep -oP '(?<=What is )[^?]+') 7 $ expr $(wget -qO- https://gajim.org:5281/register_web | grep -oP '(?<=What is )[^?]+') 5 $ expr $(wget -qO- https://gajim.org:5281/register_web | grep -oP '(?<=What is )[^?]+') 3 $ expr $(wget -qO- https://gajim.org:5281/register_web | grep -oP '(?<=What is )[^?]+') 3 $ expr $(wget -qO- https://gajim.org:5281/register_web | grep -oP '(?<=What is )[^?]+') 7 $ expr $(wget -qO- https://gajim.org:5281/register_web | grep -oP '(?<=What is )[^?]+') 15

  82. rion

    but I love this captcha anyway :)

  83. Ge0rG

    lovetox: twypnt8xlmk5ok7r@gajim.org u0a6v@gajim.org

  84. lovetox

    thanks, forwarded

  85. Ge0rG

    https://pastebin.com/Ejqsc54U

  86. lovetox

    Gajim makes XMPP great again

  87. lovetox

    user registrations just in May

  88. lovetox

    https://lagaule.org/munin/static/dynazoom.html?cgiurl_graph=/munin-cgi/munin-cgi-graph&plugin_name=localdomain/localhost.localdomain/prosody_users&size_x=800&size_y=400&start_epoch=1525803610&stop_epoch=1560363610

  89. rion

    =)

  90. Zash

    Why have open registration?

  91. Alex

    I use the simple match question plugin in Psi for a while for subscriptions. Bots are smart enough to answer the match question for a while now

  92. lovetox

    does jabber.ru have some fork of ejabberd

  93. lovetox

    they report version 3.2.x

  94. lovetox

    or is it that old

  95. lovetox

    they suffer from an older bug i reported to ejabberd

  96. lovetox

    not respecting the stream xml:lang

  97. Holger

    That's an old version of the Business Edition.

  98. Holger

    Which is based on even older code of the Community Edition.