On Oct 18, 2018, at 21:57, Máirín Duffy <mairin@linuxgrrl.com> wrote:

I'm Mo, I haven't worked on Hyperkitty in a long while but have done a lot of UX work on it in the past.

Due to various circumstances I am interested in coming back to the project and contributing again. I was hoping to maybe be able to create a new subproject on gitlab for UX issues (I use this model of doing UX work under a separate subproject and using the repo to host UX mockups for other projects and it works well.)

If that sounds reasonable and you'd have me, can you help set me up with a UX subproject? My gitlab username is mairin.

I’m not running the project any more, but it would be great to have you back Mo!

-Barry

Máirín Duffy writes:

If that sounds reasonable and you'd have me, can you help set me up with a UX subproject? My gitlab username is mairin.

I don't have a lot of time, and subprojects are Abhilash's business, but I'd like to help.

I agree with you and others that it's important. Like you, I suspect that a move to Discourse will encourage the manufacturers of heated gas and cause people who I've seen do actual work on the mailing list to drop out in droves.

But the first order of business if you're serious about keeping mailing lists as the core means of communication is not to address the forum-lovers' RFEs. It's to get the system working as designed. The Postorius part of the Mailman installation is a mess. Intermittently I'm getting 503s because the backend is unavailable. The welcome pages for the lists say signup is closed, but actually you can sign up by mail or by OpenAuth. The configuration pages don't show current settings and the explanations are inaccurate. It's pretty clear that nobody's taking care of business there. It's going to be really hard to resist calls from both top and bottom to move to something that appears to be more polished *and apparently requires little effort from Fedora admins*, unless Mailman appears to be stable and the existing UX is consistent and informative. The "manage lists" label for the link to Postorius in HyperKitty suggests an admin function or perhaps a filtering function in HyperKitty rather than a link to user configuration.

I notice that Matthew Miller complained that there are no votes via HyperKitty. That doesn't surprise me. Mailing list users have different conventions for "voting" (I'm not sure about the semantics intended for Fedora lists); they don't go to the archive interface for that purpose, and it's not uncommon for a reply to be treated as a ballot with several issues, each receiving "votes". I recall being disgusted by the idea of message- or thread-level voting in devel lists archives, but the HyperKitty developers (Fedora or Red Hat folk) insisted, so I suspect that really was driven by perceived needs of the Fedora project.

Anyway, the voting database is broken. Note the "no votes" notation and compare with the first thread in the attatched screenshot, or see http://turnbull.sk.tsukuba.ac.jp/outgoing/fedora-devel.png if it gets discarded by our mailing list. I don't have time to check the tracker to see if this has been reported/fixed in our distribution. If one purpose of HyperKitty was to allow "votes" to help decision-making, HyperKitty is clearly a failure in that regard.

I have to agree with Gerald Cox that it's unlikely that HyperKitty will ever be close to as good a Discourse as Discourse. (I also agree that Discourse won't ever be as inclusive a medium for development discussion as email is, but that's a different point.)

While I would really welcome your help in improving Mailman 3, Máirín, and the preparation of creating a UX subproject by itself is worth the effort, I think we should figure out what the real requirements are before doing a lot of coding. Up to the point you'll be happy if your effort improves Mailman but isn't implemented in Fedora, anyway.

I note that there's a similar movement on the Python core lists, and there have been a couple of abortive attempts at instantiating alternatives to Mailman (Discourse, Zulip at least) initiated from within the core developer group as well. My personal assessment is that there's very little sense that those media are a big improvement over the lists even on the tiny scale that the experiments implement, and that the real problems with Python's python-ideas and python-dev lists are pure volume of traffic and poor socialization of new posters (and a few experienced ones) into "Python list culture". I see no significant likelihood that Discourse will resolve such problems for the Python lists, in fact I believe it will make them worse by attracting more less-attached contributors, with higher discussion/ accepted code ratios. (Note, I'm one of those: I have a couple of accepted PEP coauthorships but only one 2yo PR pending. Takes one to know one! ;-) OTOH, I've learned to mostly keep my replies on my own drive, unsent, too. :-)

I suspect that the Discourse proponents on Fedora will not be satisfied with anything less than a wholesale movement of traffic to Discourse, as an experiment. I'm not unsympathetic, after all, my point is that the most any small-scale experiment can tell us is that Discourse really sucks, and I can't imagine proponents would admit that unless long-time contributors publicly complained and resigned en masse. And I doubt it's true, anyway.

Another way to say it is their diagnosis is that there's a technology problem in convenience of operation, while mine is that there's a social problem, based on human limitations in processing content and a change in customs (eg, I see a lot of configuration questions that really don't belong on a devel list -- isn't there an appropriate place for users to ask those questions?) I'm sure that a lot of posters (the ones left after the move to Discourse) will consider it an improvement, but the metric *should* be something relating to quality and quantity of improvements to Fedora provided by the list. Unfortunately I don't know how to measure that, while voting is easy.

Steve