XSF Discussion - 2019-08-28


  1. eevvoor

    linkmauve, xmpp:berlin-meetup@conference.conversations.im?join

  2. eevvoor

    perhaps a dinner with one of the berlin xmpp guys? ;)

  3. linkmauve

    eevvoor, sure!

  4. Guus

    PEP defines that a service must support the 'owner' and 'subscriber' affiliations. 0060 doesn't define an affiliation named 'subscriber'. How should I read that?

  5. MattJ

    I quietly ignored that

  6. MattJ

    Well, not too quietly actually

  7. MattJ

    https://mail.jabber.org/pipermail/standards/2018-August/035319.html

  8. Zash

    Was it removed?

  9. MattJ

    If you read it in a certain way, I said I'd submit a PR

  10. Zash

    Did you?

  11. Guus

    it wasn't removed, because I just found the same thing šŸ™‚

  12. MattJ

    I don't recall submitting a PR

  13. MattJ

    But the rules clearly state that now Guus has found it, it's his responsibility

  14. Guus

    ah, yes.

  15. Guus

    the infamous 'touched it last' rule

  16. Guus

    there's something to be said for requiring a affiliation that grants the subscribe privilege though

  17. Guus

    gotta feed the offsping - ttyl

  18. MattJ

    Guus, yes, maybe, perhaps - but it's controlled by the access_model

  19. MattJ

    So as far as design goes, there would be the choice between modelling "entity may subscribe" as either a new affiliation, or as a property of the node

  20. flow

    It really would be nice if we had an tracker for such issues (and ideally not github)

  21. ralphm

    I'm fairly sure there never was a subscriber affiliation.

  22. ralphm

    flow: agreed. Maybe we should look into hosting a Gitlab instance, but first things first. MattJ is still making an inventory for iteam, and then they can define what to do next, and ask for help if needed.

  23. Ge0rG

    what would be the principal difference between github and gitlab?

  24. jonasā€™

    one is self hosted

  25. jonasā€™

    gitlab requires some serious amount of resources though

  26. Zash

    Write them down in the wiki?

  27. MattJ

    Yeah, I'd be a -1 to self-hosting anything, for now at least

  28. MattJ

    We *had* a self-hosted tracker in the past, fwiw

  29. Ge0rG

    jonasā€™: besides of the obvious hosting question, what's the _principal_ difference?

  30. Ge0rG

    Both provide an issue tracker, so if we need an issue tracker, why not make use of our XSF github?

  31. Ge0rG

    If we need something different from an issue tracker, why use gitlab?

  32. pep.

    Ge0rG: free software/not. The feature set is more or less similar tbh. Gitlab seems to be a lot more open to big projects out there though

  33. Zash cries over Github account being mandatory

  34. jonasā€™

    pep., gitlab isnā€™t free software once you have an edition which doesnā€™t suck ;)

  35. pep.

    Gitlab's email handling is pretty cool nowadays, (I rarely need to open the browser), bit that's not something github is incapable of doing.

  36. Zash

    What were the requirements again?

  37. pep.

    jonasā€™: what features do you require for it not to suck?

  38. Zash

    What problem are we solving?

  39. jonasā€™

    properly cross-referencing issues

  40. jonasā€™

    you can link issues, but they donā€™t show up as related. you need something beyond CE for that

  41. pep.

    Zash: no idea. Somebody asked questions about github and gitlab

  42. jonasā€™

    (donā€™t confuse this with the feature of related merge requests, which *is* in CE)

  43. Ge0rG

    I don't understand how "it's not github" is a relevant feature request, considering the state of our infra.

  44. pep.

    jonasā€™: yeah I get it

  45. pep.

    jonasā€™: it's not impossible that if a big project whines a bit to gitlab they backport it to CE :p

  46. pep.

    flow: what did you have in mind?

  47. Zash

    I don't want "not github", I want "not requiring an account with a single vendor"

  48. Ge0rG

    Zash: you can login to github with your gmail! šŸ˜

  49. Alex

    Memberbot is still online until our member meeting tomorrow. In case you have not voted yet ;-)

  50. Ge0rG

    Alex: šŸ‘

  51. Zash

    !

  52. Ge0rG

    And Link Mauve has reapplied!

  53. pep.

    But he hasn't voted yet and his server is down. And mom-support doesn't know how to bring it back up :p

  54. Alex

    I accept votes also be Email

  55. pep.

    His email server is at the same place :p

  56. pep.

    Anyway, that's his issue to solve. He can contact you if he needs to

  57. pep.

    He's got jabberfr temporarily

  58. Ge0rG

    pep.: but how do we know it's actually him?

  59. pep.

    I was with him until yesterday

  60. pep.

    Or the day before

  61. pep.

    How do you know, I don't know

  62. pep.

    I was actually going to ask if that was something board and council made sure of at every session :p

  63. Ge0rG

    On some days I don't know whether it's me or not

  64. pep.

    If council members can see jids in the room for example

  65. Ge0rG

    pep.: there are questions that must not be asked

  66. pep.

    He he he

  67. Zash

    Prosody supports nickname reservations, so that could be used

  68. ralphm

    I suppose our voting shots are long enough to cover issues like this. One should vote as early as one can. And missing one vote is not a disaster.

  69. ralphm

    Slots

  70. pep.

    ralphm: agreed

  71. Zash

    But, people.

  72. flow

    pep., gitlab and discourse

  73. pep.

    Discourse is no issue tracker though. It's a forum? A place you can have discussions that are stored and that you can search through. We have the mailing lists for that

  74. ralphm

    Well, it is up to Alex to make a determination. And some things don't have to be solved by technology.

  75. flow

    pep., right, it's a forum and, depending on how you look at it, a mailing list

  76. ralphm

    On GitHub, we choose that for convenience and you don't have to use it if you don't want to.

  77. ralphm

    Primarily discussions are on our mailinglists, some here. If you want to provide new stuff or patch existing stuff you can git send-mail it, or something.

  78. ralphm

    If we indeed want to move to an issue tracker, which was decided against last time around, we can still use Github (for setup convenience) or self hosted Gitlab or something, but that will require ongoing maintenance, and people to (want to) do that.

  79. pep.

    flow: when you say gitlab is it specifically the self-hosted version, or gitlab.com would do?

  80. flow

    pep., I am pro-self hosting

  81. ralphm

    I don't see a point in moving from one hosted thing to another.

  82. flow

    err "pro self-hosting"

  83. pep.

    flow: me too generally.

  84. ralphm

    I like pro-self hosting, too. I call that my website.

  85. pep.

    ralphm: it depends. It's pretty subjective but I trust gitlab folks more than github. For one they're not entirely being hypocrites promoting free software and not using it themselves

  86. flow

    I think we miss a big opportunity by not providing source.xmpp.org and discuss.xmpp.org

  87. ralphm

    pep.: Yet?

  88. pep.

    What does that mean?

  89. pep.

    Ah, you mean they'll inevitably turn into github?

  90. ralphm

    Maybe

  91. pep.

    Sure. It's a for profit model after all.

  92. ralphm

    I don't know either of them personally and companies change

  93. pep.

    I wouldn't mind pairing with another foundation if that's a thing, tbh

  94. moparisthebest

    you might want to look at https://gitea.io/ before gitlab

  95. pep.

    For the infra

  96. pep.

    moparisthebest: yeah I had that in mind

  97. moparisthebest

    for self hosting anyway

  98. pep.

    Gitlab is a beast

  99. moparisthebest

    yea... it really is

  100. pep.

    I had to maintain one until not so long ago

  101. flow

    not saying that we could not look into alternatives, but gitlab has the most traction,

  102. flow

    and that is usually the crucial point if you are going long term

  103. ralphm

    This

  104. pep.

    I don't think gitlab is a key element tbh. I also do prefer self-host, but gitlab really is a monster.

  105. moparisthebest

    on the other hand, I think gitlab has a trello clone thing built in, so you could ditch trello too

  106. pep.

    Aaaannd we need people to support it anyway, so I suggest we leave the discussion here :p

  107. ralphm

    pep.: if you mean in that has many moving parts, sure. But it is relatively straight-forward to setup and manage, as they have this omnipackage.

  108. moparisthebest

    but as far as "how easy it is to host/maintain" and "how many resources it requires" gitea wins easily over gitlab, if those aren't your concerns, then choose based on something else

  109. ralphm

    pep.: exactly

  110. pep.

    ralphm: the omni thing is going away aiui

  111. Zash

    issue-tracker! hgweb! ... clone the prosody.im setup!! šŸ˜€

  112. ralphm

    moparisthebest: I'm not familiar with gitea (at all)

  113. ralphm

    pep.: hm, well, if they then switch to proper .debs, that works, too

  114. Zash

    Gitea seems fine. Annoys me less than Gitlab. But what was the problem being solved again?

  115. pep.

    ralphm: You mean the usual outdated .deb? With runners that are also a few years old? :)

  116. ralphm

    Zash: someone suggested wanting tickets

  117. moparisthebest

    gitea is a single static go binary, you run it, you *can* point it at a database or just use sqlite, that's it

  118. ralphm

    pep.: I don't know, many teams seem capable of having PPAs

  119. pep.

    Debian runs their infra on gitlab, but when I was maintaining it, they weren't updating the package, they were using upstream directly from what I understand

  120. Zash

    The issue was with ml threads like "should we change x?" - "yes" ... (nothing happens)

  121. Zash

    I think?

  122. ralphm

    Zash: yes, episode n+1

  123. ralphm

    Also, last time we actually said no, I believe

  124. pep.

    One thing I'd like to have is a proper search feature for our MLs, then probably the discourse song would fade a bit

  125. flow

    pep., searching is not a problem. But I really want to tag threads

  126. ralphm

    pep.: I agree that would be nice. I just seem to not delete any mail from my XMPP related mailing list folders.

  127. ralphm

    I don't know what 'tagging threads' means.

  128. pep.

    ralphm: sure, and that requires you to be on-list when the email is sent

  129. flow

    imagine tags like 'stream-management' which would show you all threads about xep198

  130. pep.

    People joining in later can't do that

  131. ralphm

    pep.: or download the mbox file

  132. pep.

    Haha

  133. pep.

    Nice ux

  134. Zash

    That would be nice

  135. Zash

    IIRC there's somewhere you can connect via IMAP and get read-only access to all IETF mailing list archives

  136. pep.

    Nice

  137. Zash

    The MUA I tried it with did not handle it well

  138. ralphm

    This should work: https://mail.jabber.org/mailman/private/summit.mbox/summit.mbox

  139. pep.

    flow: I agree that would be nice. Though generally it's already been done, people out a xep number in the subject and it works fine

  140. pep.

    (Or whatever other keywords)

  141. flow

    "ā€¦works fine" ā† how do you know that?

  142. pep.

    Because I also search in there fairly often

  143. flow

    but how you do know that you get all threads?

  144. moparisthebest

    council vote emails don't have XEP numbers in subjects, just as one example

  145. pep.

    flow: who knows.. How do you know it works anyway..

  146. pep.

    moparisthebest: I can search through that as well. Xep/features are generally mentioned in the body at least

  147. pep.

    Anyway.. It's far from perfect, but it works

  148. moparisthebest

    for some definition of works

  149. moparisthebest

    what do we need XMPP for? we can just send text to each other over telnet, it works

  150. ralphm

    Heh, standards.mbox is only 36377 messages, 192MB

  151. ralphm

    Going back all the way to July 2001.

  152. pep.

    moparisthebest: heh, you're also in the discourse camp?

  153. moparisthebest

    not necesarily, more like "downloading giant .mbox and figuring out how to use it seems like a poor way" camp haha, discourse is nice though

  154. moparisthebest

    https://ponymail.incubator.apache.org/ could be an option, I've used it here before https://lists.apache.org/list.html?dev@poi.apache.org

  155. ralphm

    moparisthebest: disappointing :/

  156. moparisthebest

    https://duckduckgo.com/?q=xep198+xep-198+site%3Ahttps%3A%2F%2Fmail.jabber.org%2Fpipermail%2Fstandards%2F that doesn't work very well

  157. ralphm

    I've briefly considered migrating to Mailman 3, but it will break the archives, or you have to keep the old archive around or somesuch. http://docs.mailman3.org/en/latest/migration.html

  158. moparisthebest

    and https://www.google.com/search?q=xep198%20xep-198%20site%3Ahttps%3A%2F%2Fmail.jabber.org%2Fpipermail%2Fstandards%2F only has 3 results

  159. pep.

    ralphm: that would be nice. We can certainly keep static archives alongside

  160. ralphm

    The are a bunch of other caveats listed there. I'm scared a bit.

  161. Ge0rG

    What about having a mirror of the archive with some other, more modern web frontend? Or nicely ask the gmane folks

  162. ralphm

    Ge0rG: oh, you mean Mailman 3?

  163. Ge0rG

    ralphm: whatever can be installed alongside the current mailman

  164. ralphm

    I'm not there's a lot of that. The problem is that Mailman 2, which we are currently running, just like probably 99% of all mailing lists, has been in maintenance mode for years, and nobody actually seems to be focussed on it anymore. Somebody would have to do the research on this, but it doesn't sound trivial. Additionally, Mailman 2 is written in Python 2, which is EOL at the end of the year.

  165. Ge0rG

    In years, or in decades?

  166. Zash

    Where are Mailman 3 sources hosted?

  167. Zash

    https://code.launchpad.net/mailman seems to suggest that Mailman 2 had commits a few weeks ago and Mailman 3 hasn't been touched in years.

  168. Ge0rG

    Reminds me of the xmpp client list activity.

  169. ralphm

    Zash: https://gitlab.com/mailman/mailman

  170. Zash

    Oh of course

  171. ralphm

    I wasn't aware somebody's actually still doing stuff on 2.1, but I don't think those are big things

  172. ralphm

    https://bazaar.launchpad.net/~mailman-coders/mailman/2.1/changes/1819?start_revid=1819

  173. Zash

    Ah, the mailman 3 link in the menu does link to gitlab

  174. ralphm

    Basically https://wiki.list.org/msapiro has been doing this on his own since 2008.

  175. ralphm

    https://mail.python.org/pipermail/mailman-users/2019-April/084332.html

  176. ralphm

    (also note that that particular list is still on 2.1 :-) )

  177. Zash

    > There are already 73 Mailman 3 lists @python.org (there are still 313 MM 2.1 lists).

  178. Zash

    Woah

  179. ralphm

    I am running Mailman myself for a couple of lists, and may attempt to migrate some.

  180. ralphm

    I was surprised how small the standards@ mbox file is.

  181. ralphm

    First snag: apparently Ubuntu packages only became available starting with 18.04, so I have to do quite some yak shaving.

  182. Ge0rG

    Get away from my yak!

  183. Zash

    Training for the 2020 Yak shaving world cup starts now!

  184. Ge0rG

    The mailman always rings twice!

  185. Ge0rG

    So I've been looking at yaxim every three minutes like a lab rat pushing the reward button, and then I realized that sending CSI is a GDPR issue

  186. moparisthebest

    Don't you just spin up another docker container in 2019

  187. Zash

    Hush now