From Meta, a Wikimedia project coordination wiki
Jump to navigation Jump to search

Internal wiki[edit]

@Guillom: Any chance you could pull that internalwiki stuff out into the public? Don't know who contributed to it so might not be possible. LuisV (WMF) (talk) 16:39, 15 January 2016 (UTC)

I don't see anything sensitive so I can probably move it. The main author at the time is now my current boss and works 1 meter from me, so asking him shouldn't be too hard :) guillom 16:51, 15 January 2016 (UTC)
Yes check.svg Done. The page is now at Discourse/2008 proposal. guillom 17:06, 15 January 2016 (UTC)


I suspect to move forward we'll need someone to volunteer their group/project/list to move. Anyone? I'd volunteer the public policy list if I still ran it, but I don't :) LuisV (WMF) (talk) 16:39, 15 January 2016 (UTC)

For your information, list admins don't own the subscribers of the respective mailing lists as a farmer owns cattle, and can't "volunteer" guinea pigs for experiments. Nemo 15:22, 20 January 2016 (UTC)

Why not both?[edit]

Could something like Mattermost or Discourse combine the dialog happening on IRC and mailing lists? Both of the proposed tools live in a space where they can be real-time (like IRC) or near-realtime (like the mailing lists). Is there value in considering bringing the worlds together?

From a technology point-of-view - one less thing to administer, maintain, etc. Also from a cultural point - less chance for cross-talking, missing out on a conversation, on-boaridng new folks, and the like.

Personally, one less place to go to be involved would be welcome. Ckoerner (talk) 16:53, 15 January 2016 (UTC)

In terms of workflows, if discourse was successful, I'm sure some communications flows could be moved into Discourse (e.g., posting some WMF news, maybe some forms of newbie help?) Maybe that achieves some of what you're thinking about? (I definitely think "fewer places to think about" would be helpful for newbies and long-term-ers alike.)
In terms of actual software, a general trend at WMF is to try to do less, not more, so we can do the remaining things better. So making discourse as "off the shelf" as possible, rather than launching ambitious new mergers of tools, is probably the order of the day. FWIW. LuisV (WMF) (talk) 17:48, 15 January 2016 (UTC)
I think we're on the same page. The gist of my suggestion is that as a community we could stop the use of two separate tools (IRC and mailing lists) and instead interact with one another in a single tool like Mattermost. I'm still learning my way here, but what does having a mailing list and an IRC channel for a given area of interest (say GLAM, or MediaWiki) give us that a single tool does not? As the movement tries to get more folks to be engaged, I think about how we have that conversation with folks. Right now our list of places to get involved is long (on-wiki,meta, IRC, mailing lists...). Half of those are things a lot of people have never heard of. Do we have an opportunity to make that easier? Ckoerner (talk) 18:09, 19 January 2016 (UTC)
My understanding is that the upstream for Mattermost hasn't made a convincing case that they have a good long-term replacement for IRC, nor that they are a viable competitor for Slack. The reductio ad absurdum argument for Mattermost: we need to do something about IRC/chat usability, Mattermost is something, we must switch to Mattermost. I personally think Conpherence is a more viable upstream provider for reasons I can discuss in a different thread.
This page isn't meant to be about IRC replacement, but rather mailing list replacement. I'm arguing for Mailman v3 because an performing an upgrade seems like a more viable iterative approach than Discourse. -- RobLa-WMF (talk) 18:27, 19 January 2016 (UTC)
Again, can someone explain what is so special about Mattermost? Last time I checked, it was just yet another web chat software with channels, like there are dozens or hundreds. It seems to also have some poorly integrated non-web applications, probably using some proprietary/non-standard protocol. Please create a table for comparison of w:Web chat software. :) Nemo 11:01, 21 January 2016 (UTC)

Notifications and new threads[edit]

(Comment by Anomie moved from the page.)

[We would lose] Useful discussion via email. Discourse reportedly supports "notifications", and maybe reply via email if configured right, but not creating new messages.[1]

I'd be surprised if the notifications are truly a replacement for a mailing list, and if the reply-by-email feature really works with inline replying and such. Anomie 17:12, 15 January 2016 (UTC)

Anomie: I think you're right about the complexity of making inline replies work well. One thing that complicated TechOps' migration from RT to Phab was email<->web integration of discussions (see Phab:T52), but it's something they were able to overcome. I'm grateful they toughed out the migration because the final unified system seems far more usable than the old diaspora of alleged "right tool for the job" solutions we used to have.
My current bias is that we seriously investigate Mailman 3, since I believe Fedora is in a similar place socially and aspirationally, and thus, we'll have a useful ally in making our online conversations a better experience for all participants (both email partisans and forum partisans). -- RobLa-WMF (talk) 18:24, 16 January 2016 (UTC)
+1 on Anomie's doubts and Fedora alliance. Nemo 10:56, 21 January 2016 (UTC)

"lose" Python?[edit]

I question the validity of listing Discourse's being implemented in Ruby versus Mailman's being implemented in Python as "a loss". I recognize we have more Pythonists than Rubyists among us, but it's not like we've been hacking on Mailman. We just used it. Ijon (talk) 21:01, 15 January 2016 (UTC)

That's probably a consequence of the page starting as a personal braindump. I think my thinking was that our current engineers and Ops team are probably more comfortable with Python than with Ruby, but your point is certainly valid. Feel free to tweak the language, and/or change the "what we win/what we lose" framing. In retrospect, we'd probably benefit from having an outline like:
  • Summary
  • Problem statement
  • Proposed solution (description, goals, risks, timeline)
  • Alternative analysis
I've already committed to other volunteer work over the week-end so anyone should feel free to edit the page boldly. Otherwise I may do it later. Guillom (talk) 21:30, 15 January 2016 (UTC)

Mailman 3[edit]

Also worth considering is the feature set of Mailman 3, which gets rid of Pipermail and replaces it with modern stuff. Fedora has mostly (I think) migrated over to it: example. Legoktm (talk) 22:51, 15 January 2016 (UTC)

Mailman 3 + Postorius + HyperKitty[edit]

My understanding is that a Mailman 3 upgrade is in our backlog (Phab:T52864) and that we're running out of excuses to delay the upgrade. The Fedora project migrated to Mailman 3 a couple of months ago (see the ambassadors mailing list for an example mailing list), and as such, it looks like a very civilized tool.

It would seem that Discourse and Mailman 3 are two candidates for solving the "make our mailing lists more civilized" problem. Is this page ("Talk:Discourse") the right place for this conversation? -- RobLa-WMF (talk) 23:14, 15 January 2016 (UTC)

RobLa-WMF: I think it's fine to discuss all possible alternatives here for now, and we can rename the page down the line if it makes sense. guillom 23:35, 15 January 2016 (UTC)

What is the problem being solved here? If it's mailman archives being an UX nightmare, upgrading will solve that. If it's the lack of collaboration features that would enable higher-quality discussions, there is no reason to expect any improvement - Mailman 3 has basically no such features apart from a like/dislike button which feels like a sad afterthought. --Tgr (talk) 01:33, 17 January 2016 (UTC)

Agreed, a better definition of the problem being solved will be important. Discourse is the idea on the table, but I believe we need to have an attempt at an objective problem statement if we are to progress this conversation. It appears that people who prefer mailing lists don't trust that their opinions will be considered.
I'm planning to further generalize my advice on this front, but I've documented the four stages of consensus building mw:Consensus. The idea of experimenting with Discourse as an alternative to Mailman v2 is clearly in the strawman phase. Who is willing to step up and lead this idea through the consensus process? -- RobLa-WMF (talk) 19:33, 19 January 2016 (UTC)

Viability of setting up a test setup on Labs[edit] is fairly nice, and we can set it up on labs easily, except for the Mail part. For that, we can use a third party mail provider to begin with (SendGrid, etc) as a starter, and migrate to our own infrastructure at some point. However, legoktm has banned me from taking on any new projects right now, so I can't do it myself but am willing to help others who want to YuviPanda (talk) 00:05, 16 January 2016 (UTC)

Collaboration features that enhance the quality of discussion[edit]

I feel like the main benefit of Discourse is being ignored here. If you want to send and read messages, there is nothing wrong with the user experience of email - in fact, it's hard to beat due to the extreme flexibility of a million different clients, some of which have been developed for costs that would equal the GDP of a small country (e.g. much more engineering and UX effort went into the toolset around Gmail than ever will into Discourse, and it clearly shows). But these tools don't focus on collaboration and constructive discussion, which is what Discourse does.

Our mailing lists have a problem. That problem is not the horrible UX of the archives (that's an annoyance, but can be worked around by tools like gmane, at least to an extent that it does not threaten the usefulness of the system). That problem is that conversation break down, trust is continuously eroded, and decisionmakers don't feel they could or should meaningfully engage. We do not incentivise insightful and respectful comments; we do not disincentivise bad arguments and outrage porn; we do not highlight the 10% or posts that express a common opinion; there are no tools to reorganize a discussion to keep it constructive or factually correct. Unlike email, Discourse sets out to do exactly this.

I'm generally cautious of automated reputation systems - they can work very well (StackOverflow, made largely by the same people as Discourse, is a masterpiece of building constructive, high-signal-low-noise communities with a permission system based on computed reputation), but it's a direction that's very alien to the Wikimedia movement (which also relies heavily on reputation, but that reputation is entirely tracked by humans, not computers). Trying to introduce software-based reputation at Wikimedia feels a bit like the Google Knol experiment for replacing (or at least supplementing) machine-based knowledge generation and assessment with human collaboration. Nothing wrong with either approach but it requires a different community DNA and one cannot just switch from one to the other and expect it to go smoothly.

That said, I think the situation is bad enough to make it worth a try. Movement-scale communication has been broken for years and shows little signs of improvement. There has been talk about the need for a Wikimedia Council / Senate / whatever that can talk for the movement. I think we just need a communication forum that works. Maybe Discourse can be that forum. Probably not, but there is nothing lost trying. --Tgr (talk) 02:08, 17 January 2016 (UTC)

I don't know about "there is nothing lost trying". There certainly could be further goodwill lost if it goes badly enough in some way. I make no prediction as to whether an attempt would go badly or not. Anomie (talk) 15:30, 17 January 2016 (UTC)
That depends on how it is delivered. If there are a few small trials, with mostly willing groups, and the outcome doesnt feel predetermined, then failure can be cheap. Sell it as the 'solution', do an enormous migration, etc, and yes, goodwill will be lost. It may be that there are some groups for which it works well, and others where it fails. John Vandenberg (talk) 13:10, 20 January 2016 (UTC)
None of the problems you mention seems to depend on software. Nemo 15:15, 20 January 2016 (UTC)
+1. Theo10011 (talk) 17:04, 20 January 2016 (UTC)

I have read these constant complaints from far too many, in far more eloquent ways. The underlying problem isn't inherent in the communication method or the software itself. Meta, mailing lists, the internet itself (from reddit to every youtube comment), has been argued as some form of a factor in breaking down the conversation. From why the trust gets eroded, and certain people don't engage to why we don't disincentives bad arguments to whatever anyone wants to blame for the status quo. I said similar things when erik decided to rename Foundation-l to Wikimedia-l as some large visionary plan for fixing our communication problems - guess what? not a single thing changed.

The problem is the people and the situations we have - they don't seem to change year after year. It's the same people commenting, disappointed in the same staff and their decisions, and it feels like the tone is accusatory and we keep repeating these cycles, blaming every problem on small things we can control, like this forum system. The entire reputation system sounds like a joke, from an encyclpoedia that was built by anonymous editors, and a culture that respects anonymity - this feels like a foolish attempt to address the real problem.

Note to staff Since the people proposing and championing this system seem to be mostly staff member from Luis, Asaf, Guillom to Yuvi etc. and this has little to no support from the wider community, as evident from the number of editors supporting this here - why don't you all try the new system internally and see what benefits there are. Personally, I am not interested in playing with a forum system. If anything this new system seems like a replacement for Meta, and I'm not interested in deprecating meta (or wiki, as a platform). Theo10011 (talk) 17:04, 20 January 2016 (UTC)

Thanks for your thoughts, Theo10011. In the hallway conversations I was able to have yesterday, I don't think there's a desire among staff to overinvest in "experiments", and I think you're right; we have many low-stakes ways of experimenting in a frugal manner among willing participants. Many of the staff involved in these conversations (e.g. Guillom) are involving themselves in this discussion out of personal interest rather than out of perceived WMF mandate (we all almost universally loathe Mailman v2 Piperman archives).
Regardless of whether we decide to experiment with Discourse, there is almost certainly going to be a need to upgrade off of Mailman v2, since it is awful to administer and the archives hate you (don't feel bad, they hate everyone). Guillom and I were both involved in the Bugzilla->something-else migrations attempts of 2010 (mw:Project management tools/2010 Review failed; wasn't ripe) and 2014 (mw:Project management tools/Review succeeded), and we agree that we need to start with a problem evaluation very similar to what we did in 2013-14 in the leadup to the Phab migration (which started life as a "why is WMF using Mingle?" discussion).
Could you offer advice on naming the problem? I admit this is bikeshedding, but I think we generally agree that "modernizing Mailman" is needed, and that even a Mailman v2->v3 upgrade is a non-trivial investment and a major UX change. I hope we can come up with a name that doesn't bias the discussion in either direction (one extreme: "mailing lists are for old curmudgeons! all the cool kids are using forums!" and the other extreme: "people who can't use mailing lists are n00bs who just need to read the FAQ!"). Thoughts? -- RobLa-WMF (talk) 18:48, 20 January 2016 (UTC)
Hey Rob. I completely agree mailman really needs an upgrade, and I say this as a frequent list user and an administrator for a local list. I really don't think you'll get any pushback in upgrading Mailman - I am all for it and happy to help any way I can. From a look at the phabricator it seems the only people resisting an upgrade were a couple of staff members, the community members are actually the one asking for this upgrade repeatedly. I don't know what risk is there, we can try it on a trial basis on a couple of lists, and revert if there are any problems - This is a task that needs some boldness and initiative.
As for our larger communication needs, I don't know about this discourse. If Yuvi or Guillom want to set up a couple of options and try new things - it would be fine I suppose but you are unlikely to get a large number of users to participate. I would guess the biggest barrier of entry to even trying something new would be - something that takes in to account the existing infrastructure and there is ease of use and lastly, visibility. So, in this case, using our existing email account and unified login would be a big benefit. At the same time, we don't want to replace Meta - perhaps something that can be added or resides on meta would be better. I don't know if such a thing even exists or we need a custom solution? Either way, this needs a carefully planned and researched approach about our options instead of let's try this cool new forum software I saw, away from meta and our email. Regards. Theo10011 (talk) 07:59, 21 January 2016 (UTC)


Are we seriously trying to make a discourse vs. mailman comparison or is this page just a joke? Surely there must be a non-biased comparison somewhere, there are hundreds of features to consider and we'd better not waste time redoing a comparison from scratch. Either way, I think a discussion can't even start without a clear problem statement and ideas on how to measure success. Nemo 15:14, 20 January 2016 (UTC)

Other forum software[edit]

Using forum software is a perennial proposal in Wikimedia since at least 2002, so forgive me if I'm already tired of discussing the "idea". Anyway, to be more productive I think we should all edit w:Comparison of Internet forum software with reliable sources for the benefit of an informed discussion among us and everyone. Personally, if we really needed to adopt some forum software I'd go for a field-tested one with a solid experience, which the users will be used to, like phpBB. Nemo 15:21, 20 January 2016 (UTC)

For that matter, if we're talking forum software there's also Flow. But this is about a mailing list replacement (supposedly Discourse is that too) rather than just adding a forum. Anomie (talk) 16:08, 20 January 2016 (UTC)
Is there some explanation on what makes Discourse, more than other forum software, a replacement for mailing lists? Is it merely the ability to reply to existing threads by email? With that criterion, even Phabricator is a replacement for mailman. :) (Note, I can't read [2] because it requires JavaScript.) Nemo 10:35, 21 January 2016 (UTC)
I've started adding some information, please help. Nemo 10:54, 21 January 2016 (UTC)

Wikimedia account support[edit]

Is there any chance that this will allow users to participate in discussions using wiki accounts, instead of only email addresses? --Yair rand (talk) 21:46, 24 January 2016 (UTC)

Yes, Discourse supports "OAuth integration with MediaWiki credentials." so we would log on to Discourse using our Wikimedia account. John Vandenberg (talk) 23:44, 24 January 2016 (UTC)
Table expanded, please add a reference. Nemo 07:23, 25 January 2016 (UTC)
Note that Mailman 3 is also designed to allow this and chose Persona as authentication method (ready or not?) and there is a call to save the Persona extension. Nemo 07:29, 25 January 2016 (UTC)
Yair rand, John Vandenberg and Nemo There is a pilot installation at "Configure Single Sign On at" or "OAuth integration with MediaWiki credentials" has been requested at Phabricator, see phab:T124691. Ad Huikeshoven (talk) 17:03, 28 January 2016 (UTC)

Quality of the email support[edit]

I have used other Discourse instances, namely for Ubuntu and the Atom editor, and would like to note that the email support is very bad. Messages don't arrive in the right order (you regularly get a message that was sent after another message before that message, which makes following discussions in an email client confusing), editing posts doesn't work, mailing list headers aren't set properly and there are many other major or minor issues. What's more important is that the email support is actually implemented as notifications and that all the messages sent by Discourse are HTML, not plain text. It's not at all a replacement for a mailing list, and I don't think the email support was given enough thought. --Rastus Vernon (talk) 03:59, 10 February 2016 (UTC)

No plain text mode, seriously? :o That sounds like a serious bug, is it filed? Nemo 21:13, 10 February 2016 (UTC)
I don't think it'd be seen as a bug, it's just a major shortcoming. Another such major feature lacking is threading. There is no way to thread discussions as can be done in mailing lists, which is another dealbreaker. --Rastus Vernon (talk) 04:25, 19 March 2016 (UTC)
I think the no-threads thing is seen as a feature of Discourse. See this by one of the founders of Discourse. Personally I rather like it; I always find hierarchical threaded conversations harder to follow than linear ones (especially as then people are forced to clarify what exactly they're replying to). — Sam Wilson ( TalkContribs ) … 05:52, 23 March 2016 (UTC)
You can't really "force" people. Threading is essential to survive with conversations such as wikimedia-l, e.g. to (un)follow (un)interesting subthreads. Nemo 07:07, 23 March 2016 (UTC)
No, sorry, you're quite right; I meant that people do rather clarify their meanings when they're replying out of context. It sometimes makes for more verbosity though. — Sam Wilson ( TalkContribs ) … 00:25, 29 March 2016 (UTC)
I'd be curious how many people use a threading email client these days; surely most use gmail with its flat views. If wikimedia-l is considered "unsurvivable" without features unavailable to most of it readers (not to mention most would-be readers in the Wikimedia community), that would be a pretty strong argument against the status quo. --Tgr (WMF) (talk) 08:47, 29 March 2016 (UTC)
Currently, people are free to choose between efficient and inefficient tools. Of course we can't expect everyone to be efficient. People who use inefficient tools exhibit various issues in their usage of the lists, but I won't go into personal details. Nemo 11:41, 29 March 2016 (UTC)

I found out that FUDforum has a native NNTP gateway, and Mailman can also be configured for a NNTP<->mail gateway. Perhaps we should be testing FUDforum rather than Discourse, if mailing lists are your target. Nemo 07:02, 24 March 2016 (UTC)

discourse unusable[edit]

discourse has too many weak points imo:

  1. it is not a proper mailing list server. it e.g. reformats mails, cuts out email adresses etc
  2. it is not clear why this should be better than a mailing list
  3. the mobile interface has a small font and big everything else, eating up screen space
  4. resizing text in mobile is not possible
  5. it is slow, and an additional web page
  6. the main point is that one wants to send the same mail onto different lists. and the answer is visible also even if you are not subscribed. but - linux mailing lists are doing this for ages. they have a no subscription model. they have mailing list archives perfectly searchable.

to sum up: if you want to reach people, reach them where the are already, do not invent something new which is worse than everything else used. something like facebook, posting, liking, voting, etc would be nice, but this should run on wikinews. if we want to improve mailing lists, expanding mailman seems much more appropriate. --ThurnerRupert (talk) 18:47, 19 March 2016 (UTC)

Not a fan of mailing lists, but another idea is Ben Creasy (talk) 22:38, 1 April 2018 (UTC)

Site online again[edit] is back online. Sam Wilson 05:11, 20 November 2017 (UTC)

Thanks for the update. --Nemo 11:04, 20 November 2017 (UTC)

Testing the path from mailing lists to Wikimedia Space[edit]

Hi, I have just published a proposal to start testing mailing list mirroring and emulation at the Discourse instance we have at Wikimedia Space. If you are interested in discussing and exploring potential post-Mailman scenarios, join in.

Check the proposal. Feedback is welcome there or here. :)

If you want to propose a mailing list to mirror or you are personally interested in testing mailing list emulation, let's talk. Qgil-WMF (talk) 21:23, 29 October 2019 (UTC)

(FYI all the questions received in different channels are being collected at the proposal's discussion.)
What_we’d_lose_(user_experience) lists some points as potential disadvantages of Discourse over Mailman.
  • Problems with email rendering? The promise is to offer no functionality loss to Mailman users. We want to test mailing list mirroring and emulations precisely to get a first impression about problems in real use if any (and file bugs accordingly).
  • Email forwarding? Same thing, let’s test and see what are the problems if any.
  • Problems subscribing. OK, there is room for improvement indeed, both for making easier to understand how subscribing works (the subscription itself is easy, clicking a button) and for implementing mailing list mode per category. Again, testing just helps understanding the actual impact of this problem (which can be compared with the problems that users face the first time they learn about mailing lists, which also brings a learning curve).
Other points are listed in that section but I don’t believe they are relevant, or actual “losses”. Qgil-WMF (talk) 23:22, 30 October 2019 (UTC)