Stream: project-inline-asm

Topic: rfc thread


nikomatsakis (Jan 22 2020 at 16:08, on Zulip):

Hey -- I gave the RFC (rfc#2850) a quick read this morning. Overall, I'm quite positive, but I'd like to make a meta-note: this rfc thread is getting really long! So far it strikes me as mostly some high quality questions and answers.

I'm wondering though best to manage it. Perhaps things are just fine here, but one idea I've been toying with for longer threads is to close the RFC periodically and collect the questions etc into a FAQ and summary, and then open a new RFC that starts with that summary. The hope would be that people who come to the RFC would be able to quickly pick up the context and avoid repeat questions, and that it would be RFC authors a bit of time to catch up on people's questions and integrate any changes they plan to make.

That said, it may not be necessary in this case. =)

nikomatsakis (Jan 22 2020 at 16:08, on Zulip):

I'm curious @Amanieu / @Josh Triplett what you think about this larger question (should we try to "curate" RFC thread at all, to help people get an overview of what's been discussed, and -- if so -- how?)

Amanieu (Jan 22 2020 at 16:11, on Zulip):

One idea I had was to use the project-asm issue tracker to keep track of multiple threads of discussion.

Amanieu (Jan 22 2020 at 16:11, on Zulip):

We did use it that way while we were drafting the RFC.

Josh Triplett (Jan 22 2020 at 16:14, on Zulip):

At the moment I feel like most of the threads have been resolved by updating the RFC.

nikomatsakis (Jan 22 2020 at 16:35, on Zulip):

One idea I had was to use the project-asm issue tracker to keep track of multiple threads of discussion.

yeah, that's an alternative, although maybe it's a bit late for that?

nikomatsakis (Jan 22 2020 at 16:36, on Zulip):

i.e., there's a lot of stuff on thread already :)

nikomatsakis (Jan 22 2020 at 16:36, on Zulip):

At the moment I feel like most of the threads have been resolved by updating the RFC.

OK, this wasn't obvious to me

Josh Triplett (Jan 22 2020 at 18:15, on Zulip):

@Amanieu I absolutely think we should track individual issues in the issue tracker, and use that for cases where we haven't updated the RFC yet.

Lokathor (Jan 23 2020 at 00:14, on Zulip):

Discussion living in the issue tracker for longer definitely keeps things under control better. For future tricky RFCs we might want to just immediately say "please do not discuss here, please use the issue tracker at this other place for better threaded conversation"

nikomatsakis (Feb 05 2020 at 22:54, on Zulip):

I'd like to re-open this. I feel like the conversation on rfc#2850 is way more than I can reasonably be expected to digest. :) I think it would be good to close the RFC and try to summarize key points in the discussion thread, personally.

nikomatsakis (Feb 05 2020 at 22:54, on Zulip):

That said, probably the most imminent thing is to resolve the impasse around the project group RFC first

Amanieu (Feb 06 2020 at 17:21, on Zulip):

I'm on holiday this week, but I'll try to get a summary written some time next week.

Amanieu (Feb 06 2020 at 17:22, on Zulip):

If we close the RFC thread, where should discussion continue?

simulacrum (Feb 06 2020 at 20:06, on Zulip):

@Amanieu I think the idea is that we close it and until we have a summary we basically say "please hold off on discussing"

mark-i-m (Feb 07 2020 at 23:16, on Zulip):

I think it might be good to do a retrospective on this new RFC process after it's all finished. I might be an outlier here, but I feel that it didn't reduce the amount of discussion after the RFC was published.

Lokathor (Feb 08 2020 at 03:44, on Zulip):

I think for an RFC of this nature, the RFC should be first PR'd into the project group repo and go through a round of all the small edits and oversights.

only then, once the basic errors have been hammered out, would the RFC go to be PR'd to the actual RFCs repo and get the wider audience attention.

Amanieu (Feb 09 2020 at 13:47, on Zulip):

The RFC actually went through several stages of feedback: first the pre-RFC thread on internals, then the project group and finally the published RFC.

Lokathor (Feb 09 2020 at 17:36, on Zulip):

I'm aware of all of those stages. I even participated in some of them. One Discourse thread, and then a github repo with 6/10 issues resolved, and three PRs to the files.

I maintain that it was insufficient prep work for something of this scale, and that requires as much detailed specification as this does.

Basically because discussion on a PR thread is a terrible way to manage topics and allow people to go through it later to catch up on what has been talked about. That's why it should be Zullip or GitHub Issue based for much longer.

Amanieu (Feb 09 2020 at 18:09, on Zulip):

To be fair, at the point where I decide to publish the RFC, there were no more outstanding issues on both zulip and the github repo, and discussion seemed to have somewhat died down.

HeroicKatora (Feb 09 2020 at 18:29, on Zulip):

Moving from internals to Zulip/Github left a number of open questions behind. Were all of those posed again, or left unanswered? Determining an answer to that question seems to require going through both threads manually which is a huge undertaking of its own.

Josh Triplett (Feb 13 2020 at 20:17, on Zulip):

We talked about the RFC thread in the lang team meeting today. We definitely felt that the current thread would benefit from a summary post (covering all outstanding issues), along with making sure the alternatives section of the RFC is complete with respect to things raised in the thread.

@nikomatsakis suggested, and I'm starting to agree, that due to limitations in GitHub we might want to close the RFC PR and open a new RFC PR, to draw a line under the discussion so far and help people start fresh.

Josh Triplett (Feb 13 2020 at 20:17, on Zulip):

1) Does that seem reasonable?
2) @Amanieu Are you up for writing that summary and updating the RFC accordingly?

Amanieu (Feb 13 2020 at 20:26, on Zulip):

Sure I'll do that.

Amanieu (Feb 13 2020 at 20:27, on Zulip):

By alternatives section, do you mean unresolved questions section?

Josh Triplett (Feb 13 2020 at 20:28, on Zulip):

Unresolved questions, and the Alternatives part of Rationale and Alternatives.

Josh Triplett (Feb 13 2020 at 20:32, on Zulip):

Effectively, any outstanding reasonable comment that we don't fix by changing the RFC should be documented either as an unresolved question (if we want to defer it to the implementation phase to answer) or under rationale and alternatives (if we want to say "we could do it differently, but we don't think that's the right approach, and here's why").

Josh Triplett (Feb 13 2020 at 20:32, on Zulip):

People should be able to just read the RFC, not the whole thread, and know what paths we considered and didn't go down.

Amanieu (Feb 13 2020 at 22:31, on Zulip):

@Josh Triplett incidentally, any progress on moving forward with the project group RFC and the llvm_asm! RFC?

Josh Triplett (Feb 13 2020 at 22:31, on Zulip):

Working on that now. :)

Josh Triplett (Feb 13 2020 at 22:32, on Zulip):

Trying to figure out the right next step there.

Amanieu (Feb 25 2020 at 12:08, on Zulip):

@Josh Triplett Sorry about taking so long to do this, I'm working on writing up a summary now. Do we want to encourage people to comment on a new RFC thread (open a new PR on rust-lang/rfcs) or should we point them to use the issue tracker of the asm project group instead?

Amanieu (Feb 25 2020 at 14:37, on Zulip):

I'm not exactly sure what to write for the summary though. I triple checked all comments on the RFC and made sure everything was addressed in the RFC text.

Josh Triplett (Feb 25 2020 at 15:12, on Zulip):

Then the summary should say exactly that, that all comments above this point have been Incorporated into the RFC. And you should link that summary from the top message. You should then mention that as requested you're opening a new thread to make it easier for people to follow.

Amanieu (Feb 25 2020 at 15:21, on Zulip):

OK, will do.

Amanieu (Mar 03 2020 at 22:12, on Zulip):

@Josh Triplett @nikomatsakis I didn't attend the lang team meeting where the decision to close the RFC thread and open a new one was made. Could one of you reply to this comment to explain the rationale? https://github.com/rust-lang/rfcs/pull/2850#issuecomment-594126830

Lokathor (Mar 03 2020 at 23:51, on Zulip):

I don't think that you, Amanieu, specifically did anything wrong, because you were clearly told to restart the RFC, but I think the idea to restart the RFC was a poor one (for essentially the reasons stated in that comment).

Josh Triplett (Mar 04 2020 at 00:33, on Zulip):

I'll try to put a response together, sure.

Josh Triplett (Mar 04 2020 at 00:34, on Zulip):

I do think it was a good idea, due to limitations of GitHub.

Josh Triplett (Mar 04 2020 at 00:48, on Zulip):

Done.

RalfJ (Mar 04 2020 at 17:44, on Zulip):

(FWIW I agree the closing + reopening was a good idea.)

mark-i-m (Mar 07 2020 at 23:57, on Zulip):

I personally also agree that reopening the thread was a good idea, but I'm curious if any feedback has been given to github about this... it seems like we run into it pretty often

Last update: Jun 07 2020 at 09:55UTC