(linked discussion) Ansible's migration from MLs to Discourse vs. Fedora's

The Water Cooler
1

Continuing the discussion from F46 Change Proposal: Changes Discussion Only On Devel List:

The Ansible Community SC never used mailing lists for community discussion or coordination in the way that Fedora does. We had a Github issue tracker which was (mis)used for discussion and then moved it to Discourse. Everything else Ansible-related that was moved to Discourse (read-only announcements, event announcements/calendars, user support, the mostly dormant ansible-{project,devel} lists, etc.) was either better suited to Discourse or neutral.

1 Like
2

Ah, so logically if they weren’t depending on mailing lists as heavily then it was much easier for them to transition away from them. That said, I still think the main take away should be the short transition period rather than trying to do the split brain long term. Giving Discourse a fair shot would mean actually using it exclusively for change proposals for some amount of time.

You also mention that several things were better suited (or neutral) for Discourse. That made me realize something: the change proposal as written focuses almost entirely on why split brain is bad, but to me doesn’t substantially address why email is better suited for change discussions. The replies advocating for email seem to largely focus on personal preference around clients and filter setups (extremely subjective metrics), not why email is actually better for most people. There are many Discourse features (markdown, reactions, rich quoting, polls, tags, slow mode, thread splitting, etc.) that apply equally to everyone and don’t require having years of experience wrangling email clients and filters to make usable.

3

I don’t think the length of the transition really had anything to do with it. It was largely about community sentiment and preferences and enthusiastic adoption that made it successful. The Ansible SC’s previous setup of (mis)using various Github features was nothing that anybody (including myself) was particularly drawn to. Also, I think the fact that the Ansible community is smaller has helped; these medium-large discussions that are quite common in Fedora (but not Ansible) make Discourse rather painful to use.

I also think the Ansible Forum’s approach to moderation has been helpful, but our issues are more about various bad actors creating new accounts to post spam than more-difficult-to-address conduct issues that happen here. There were proposals to adopt an AI spam detector thing, but instead we configured Discourse to require manual approval of all new accounts and expanded the moderator team, which has largely solved the problem.

4

Another data point.
The python devs voted with their feet and moved off email to discourse a while ago.
Only a few users still use the python users email list.

1 Like
5

This was actually tried for a while. At the beginning of the experiment, all discussion was directed to discourse, and posters to the mailing list were reminded to post their responses to discourse, and not to the mailing list.

It just didn’t stick. And Ben’s departure as FPgM didn’t help.

FWIW I always post my own feedback for Change proposals on discourse, and try to only interact with the mailing list when responding to something that was only raised on the mailing list. I’m trying to do the “right thing” :trade_mark: here, even though I thoroughly dislike how many discussions turn out on discourse. I’m beginning to think that Discourse as a tool would be OK, but the way we’re using it is just a bad fit and / or very light moderation increases the risk of brigading.

1 Like
6

:sob:

We can certainly add that in as a contributing factor.

Perhaps the second part was a contributing factor as well. If people who prefer email can get responses to their concerns on the mailing list, there’s no incentive to communicate on Discourse. I respect that you wanted to acknowledge and respond to people providing feedback that way, but that helps enable the split brain that is cited as the primary driver of this change. If replies on list went unanswered, or one of the other suggestions was implemented, perhaps we could have avoid the split brain. What this change is essentially saying is “it didn’t work because we didn’t fully commit to it”.

I think trying to fix that problem is a better solution than abandoning Discourse. The term “using a sledgehammer to crack a nut” comes to mind. Sure, reverting to email-only would reduce the volume, but it would be throwing out so many valued voices just to avoid hearing from a few troublemakers.

1 Like
8

EDIT: I deleted my response and resubmitted it from the Web UI. The Discourse email integration deleted half of it :frowning: .

(apologies for the wall of text; I tried to address all of your points in detail)

If replies on list went unanswered, or one of the other suggestions was implemented, perhaps we could have avoid the split brain. What this change is essentially saying is “it didn’t work because we didn’t fully commit to it”.

I disagree. The problem is more than that we didn’t discourage “reply to list” hard enough, and I don’t think that’s the proper way to go about things in the first place. A forcing mechanism was intentionally left off of the original FESCo Discourse experiment ticket. I think the multi-step strategy to only move Change discussions away from the devel list and then move everything else at some later date (which has not materialized) was flawed. @berrange summed up the issue with this aptly in the devel thread for this Change.

Excerpt:

If there is a desire to make the forum the exclusive venue for
discussing Change proposals, then IMHO that is more fundamental
than just the Change process. It would suggests a wholesale
re-evalation of the maintainer expectations & documentation for
communications. To me that would be an undesirable switch.

(For the record, that would be undesirable to me also, as things stand.)

I think trying to fix that problem is a better solution than abandoning Discourse. The term “using a sledgehammer to crack a nut” comes to mind. Sure, reverting to email-only would reduce the volume, but it would be throwing out so many valued voices just to avoid hearing from a few troublemakers.

There have been a lot more than a few troublemakers coming from Discourse. I don’t think we need to turn off the devel list to see this problem. It’s been clear from the first F40 Change that was submitted for discussion on Discourse three years ago.

And switching entirely to Discourse would throw away valued voices on the devel list, in particular the active packagers who are required by policy to follow announcements on the devel list. That’s bad also.

The bigger question is: who is going to fix the problems with Discourse if we double down and get rid of the devel list? Who is going to address workflow issues for existing email-using developers (which are not the only stakeholders but are very important ones)? Council pushed Discourse as the place for both project discussion and user support (Ask Fedora), but now nobody owns that effort or the problems it’s causing. I made my own suggestion in Consider moderating first posts from new users and previous efforts to improve moderation apparently haven’t gone anywhere either.

Therefore, defaulting back to the devel list (with the limited Discourse cross-posting and polls) arose as the most workable solution to solve the split-brain issue, and potentially the Discourse-specific issues.

the change proposal as written focuses almost entirely on why split brain is bad, but to me doesn’t substantially address why email is better suited for change discussions.

This was somewhat intentional on my part. I opted not to write a polemic on Discourse in this Change, as the issues with Discourse have already been discussed multiple times, and the most recent discussions were linked in the Change. Also, the Change Owners each have slightly distinct views about the primary issue(s) that Discourse Change discussions caused; they mainly surround moderation; conduct; and UI/workflows. Out of them all, the split brain issue is the most clear cut. It most acutely affects FESCo members, Change Owners, and the Change Wrangler and has caused delays, miscommunications, and inefficiencies in the Changes Process. We had to make a choice to solve it, and I believe it should be the devel list (at least for now).

We also discussed ways to find ways to keep the "solely stick to Discourse” users in the loop and not make things completely invisible. That was an explicit goal of mine (I understand if people think this was insufficiently addressed, but it was a strong consideration). This includes keeping individual posts on Discourse (as opposed to digests that were originally proposed) and emphasizing the polls. I had hoped that the polls would make it possible for all community voices to be heard without the brigading problem. If anybody has additional concrete, actionable steps to add to this Change to bridge the gap, please let me know.

There are many Discourse features (markdown, reactions, rich quoting, polls, tags, slow mode, thread splitting, etc.) that apply equally to everyone and don’t require having years of experience wrangling email clients and filters to make usable.

Honestly, I think most of those are anti-features that can make contentious discussions less constructive, and I still find Discourse unpleasant to deal with despite my “years of experience.”

1 Like
9

To be honest, I do not understand why removing discourse users from the conversation completely can be considered ok. But asking e-mail users to use a different reply-to address when discussing change topics is such a controversy.

Yes, it is a bit of a friction, yes there are some issues with the formatting of quotes maybe, but still out of those two the first one looks much worse to me.

2 Likes
10

(Discourse please don’t mangle my quoting; there should be one in-line
response. I’m hoping that inserting extra text here might make this work.)

EDIT: It was still broken!

there are some issues with the formatting of quotes maybe

I usually trudge by with the Discourse web composer but have been trying
to test the email interface as part of this Change. It turns out the
issues with formatting are more severe than I originally realized. It’s
not just that quotes look different since they don’t have the proper
metadata. For most of the posts on this Change thread I’ve attempted to
make via email, Discourse has mangled my message in one of (at least)
three ways: combined my response into the blockquote; deleted quoted
sections entirely; or deleted entire chunks of unquoted text that I
wrote. I keep having to go back to the webui to make sure my posts are
not mangled, and in many cases, they have been.

And quoting is obviously not the only issue people are facing. I don’t
understand why the onus is on existing contributors who are already
stretched thin to adapt to tooling that they find sub-optimal,
stress/anxiety-inducing (multiple contributors), “confusing and
irritating” (quote from a previous thread), and otherwise unpleasant to
deal with. It feels like we’ve set up this strawman that the main thing
stopping new people from getting involved in the project is that we’re
not using Discourse, and I don’t think it makes sense. Nor does “why
can’t you just respond to Discourse by email; it’s basically just a
mailing list with a web UI.”

11

That’s not the argument I am making.

You are proposing to completely remove the ability for Discourse users to participate in the Change discussion.

While you are not willing to consider the non-blocking inconvenience for the e-email users to participate in it.

(And yes, every time I write a mail to devel@, after hitting the send button in the UI, i have go to Hyperkitty interface to make sure that it was delivered correctly, not lost in some moderation queue and the formatting is not broken. I have been dealing with this non-blocking inconvenience for years without complaining. )

1 Like
12

That ticket explicitly said that “discussion will be directed” to Discourse. It didn’t specify the exact enforcement mechanism, but clearly indicated that there should be one, also confirmed by Matthew’s follow up post with ideas about what that should look like. Just saying it was “intentionally left off” misrepresents the history. Let’s figure out why FESCo didn’t figure out the enforcement mechanism instead, and then try actually enforcing it.

Sounds great to me, let’s move all Fedora development discussions to Discourse. The split brain is caused by people refusing to give up email, no matter where the line is drawn. It’s deeply ironic that a project known for innovation is being held back to a 55 year old technology because a minority of the project refuses to change.

And there are lot more than a few troublemakers on email. The difference is on Discourse we have real moderation tools to deal with them.

And switching entirely to email would throw away valued voiced here on Discourse. But these choices aren’t equal, as email-favoring people are still regularly replying here in Discourse, but Discourse-favoring people are almost exclusively not replying on email. One of these is objectively more damaging to participation in the project.

All of us. Right now I’m doing my part by starting a discussion in the mod coordination category about the feasibility of restricting conversations in change proposal threads to either FPCA+1 or a minimum Discourse trust level. Another idea that came up in that thread was defaulting change proposal threads to slow mode.

In the spirit of open source, this is a “scratch your own itch” problem. People using the website have none of those issues, and it’s not their responsibility to fix an integration they don’t use. People that care about the email bridging problems should work together and consolidate their feedback to the Discourse developers, routed through the Fedora contract if necessary to give the feedback more weight. Or they can even take a stab at sending a pull request to fix it themselves if they are so inclined. The obvious counter argument is who is going to address the fundamental issues with email itself that make it so unappealing to the majority of people? Unlike Discourse, there isn’t a path to improving it. I’m sorry, but I’m completely done with “you’re holding it wrong”-type arguments about clients and filters. Email is bad, and just because some people figure out a way to make it work for them does not mean that it is a good choice for the whole project.

Who owns the effort to direct discussion away from email and towards Discourse, which was half of the original approved proposal?

Not getting your way with one suggestion does not mean that moderation is not improving.

I appreciate the consideration, but it misses the point. The problem is not visibility, people can read the wiki for that, or even subscribe to devel-announce (i.e. single announce emails versus long threads). The problem is being able to provide feedback and participate in the discussion.

  • markdown: Objectively beneficial, not destructive.
  • reactions: Very beneficial. It’s so much easier to react with a :+1: or :heart: than to send an email of just “+1” or “I agree” (which ironically makes the volume of email much worse when people do that). We don’t even have negative reactions like :-1:.
  • rich quoting: Objectively beneficial again. Email quoting is extremely fragile and inconsistent across clients. Meanwhile above I was able to attribute a quote you shared from email directly to Daniel to not make it appear like it was your statement.
  • polls: Since you want to emphasize these with the change, I’ll assume you also see how these are valuable.
  • tags: Objectively beneficial, not destructive.
  • slow mode: Critically important moderation tool, with no equivalent on the email side. When email threads get out of control (which does happen, and this change won’t fix), people just abandon keeping up with them, because there is no slow mode. The only arguments against slow mode I’ve seen are people complaining about being silenced or censored, which we can mitigate by starting change threads in slow mode so no one feels like it was in response to their post.
  • thread splitting: Another critically important moderation tool, with no equivalent on the email side. Some people complain about their post being moved off to the side, but that is easily debunked by the fact that there is a link to the split in the main thread, and if anything it elevates their split discussion to the same level as the change thread itself instead of being buried inside the main thread.

I honestly can’t see how any of these can be argued to be anti-features. The closest would be reactions, if we had negative reactions like :-1:, but we don’t.

4 Likes
13

Note also that we can address specific email integrations issues, once they are reported.

For example, let’s say we do not find a good workaround for quote formatting issue[1], then we can just literally add a rule that moderators can fix the formatting of the post as long as it preserves the meaning.

Or we find there is an issue with formatting tables - same story. We report it, we discuss it, we find a workaround.

We basically can say “we prioritize the availability of the Change Discussion category over E-mail. The moment someone posts something incompatible with the email workflow, we set up a flag to notify moderators to address it in the way we agreed beforehand”

  1. since @kevin uses email workflow quite regularly I would assume there is a workaround ↩︎

1 Like
14

I don’t think it’s fair to put this fully on FESCo’s shoulders.

The decision to do discussions on discourse was approved on May 9, 2023:
https://forge.fedoraproject.org/fesco/tickets/issues/2989#issuecomment-835074

And on May 12, 2023 (only three days later!), Ben Cotton announced on his blog that he’d been let go from Red Hat.

I don’t think FESCo can be blamed for the chaos that ensued around Change proposals in the months that followed.

1 Like
16

I’ve acknowledged previously that Ben getting laid off was a contributing factor. It wasn’t the only factor, or even the primary factor. I don’t think anyone expected the FPM to solely be responsible for the email-to-Discourse redirection, so losing the FPM can’t be the sole reason. On the contrary, I think it was many small interactions that weren’t to blame individually, but combined created the environment and culture where it was acceptable to continue posting change feedback on the mailing list. Things like change owners commenting on their change and asking people to send feedback to the mailing list (literally the opposite redirection), or replying to feedback on the list instead of redirecting it to Discourse. There is shared blame to go around, it’s not solely on FESCo, but FESCo is not excluded. When I ask about the why, I’m asking if there are any other significant events we need to factor in and be conscious of when trying again. It’s not about blame for the sake of blame, it’s about learning and improving.

19

EDIT: Discourse email integration ate half of my last response; reposted it

Not getting your way with one suggestion does not mean that moderation
is not improving.

This is a rather uncharitable interpretation of what I said. My point was
that nobody’s efforts to improve moderation of Changes have appeared to
make a meaningful difference. You mentioned there’s a new discussion
about instituting additional discussion requirements for Changes. Can you
provide a link?

I’m sorry, but I’m completely done with “you’re holding
it wrong”-type arguments about clients and filters.

Did you read the prior discussions linked in the Feedback section of the
proposal? This exact point in reverse (“being repeatedly told that I’m
just holding discourse wrong is infuriating”), and other similar ones,
were raised in the Matrix chat and elsewhere. I don’t think forcing
mailing list users to use Discourse and saying, look, some of them tried
to use it, is particularly compelling, especially considering that some
of them later burned out on Discourse.

(deleted my off-topic commentary about the Matrix transition vs.
attempted Discourse transition)

Re. historical context, I’ll note that creating Discourse Change threads
and figuring how to “steer” people toward them from the ML announcements
made up the first step of a multi-stage Council plan to sundown mailing
lists that a Council member announced on devel@ in 2023. Staffing
changes in and around Council and associated burnout definitely did
impact this work, and the plan was never fully implemented. FESCo was
just asked to sign off on that first step. Going forward, I do think
FESCo should take a larger role. I would much rather discuss what we can
do for the future as opposed to engaging in a blame game.

Anyways, perhaps we should agree to disagree here — circular
discussions are no fun. I understand that we have very different
perspectives about Discourse and email and mailing lists :slight_smile: .