Hi,
I personally don’t really have any particular opinion on moving to Discourse, versus staying on mailing lists (if pushed, my naturally conservative mindset would say stay/use mailman 3 as discussed before, but I’d probably adapt to a switch quickly enough). However, I do have some related concerns to do with the process in particular:
- Regarding this paragraph in the blog:
“The majority of the community was in favor of the move when the move to Discourse was discussed extensively on the LLVM mailing lists. This provides the features mentioned above in addition to a more modern communication. We did hear of one feature some would miss compared to Mailman: the ability to reply to someone directly through email. However, while it may not be ideal for some, we feel this is a worthwhile tradeoff to gain the other benefits, e.g. better safety for LLVM developers and users in general.”
I skimmed the most recent thread on this topic from the middle of last year, and the distinct impression I got was that the majority opinion, or at least about half of those posting were actually against any move to Discourse, with several raised concerns that I never saw addressed (topics about accessibility and disagreements from existing moderators to the point about moderation being a problem on mailman being two examples). I haven’t gone over the thread which originally introduced Discourse back in 2019, so I can’t say what happened there. Was this majority reached in the 2019 thread, because my memory of it was that there was no clear consensus in either direction?
-
Also from the above paragraph: who is “we” in “we feel this is a worthwhile tradeoff”? If referring to a specific subgroup (e.g. the IWG/the board), were these concerns actually discussed with the people who raised the concerns? If not, this seems to me like a case of “others don’t agree with us, but we’re going to ignore their concerns and go ahead with what we (i.e. the IWG/the board etc) want to do” which isn’t how community consensus works…
-
The category structure: “January 7-9 - Re-configure the existing LLVM Discourse to the new category/subcategory structure (see below)”
When was this structure discussed? Note that the mailing list announcement came AFTER this point of time had started, meaning there was zero opportunity for people like myself who have concerns with the category breakdown to raise them and suggest improvements. Contrast this with the Github Issues migration, where I was able to get additional categories added to the list of labels, to reflect the pre-existing bugzilla breakdown, and how I used this.
Three particular categories of topics that aren’t reflected in the breakdown are a) debug information, b) LLVM tools like llvm-readelf, llvm-objdump, yaml2obj etc, c) testing infrastructure, i.e. lit, FileCheck etc.
- The timeline: “January 10-20 (sometime during these 2 weeks) - The LLVM mailing list archives are migrated to Discourse and it is sanity checked by volunteers of the LLVM community. This sanity check can take a week or more.” and “We encourage all LLVM community members to start using Discourse on Jan 10th to minimize any disruption once the mailing lists become read only and the final messages are merged to Discourse”
Given that this timeline starts today, and was only announced over the weekend (my time), there is zero opportunity for anybody to raise concerns or points, made worse by the fact that many community members might be off for a couple of weeks without any idea this is going on. This timeline should have been at a minimum 2-3 weeks after announcing it before it even begins. Again, contrast this with the recent bugzilla migration, where there were plenty of opportunities for others to raise feedback, and time to address them, before the migration even started. The 1st of February is the earliest any of this should have been starting, in my opinion, not the final cut-off!
James