Stream: t-compiler/wg-meta

Topic: MCP procedure


nikomatsakis (Jun 04 2020 at 15:14, on Zulip):

So @Santiago Pastorino there are two things I don't like about MCPs "as practiced today".

The first is that we often put the design in an "out of line hackmd or gist", and the template is maybe too wordy. I find it's kind of hard to find the actual proposal in there, it's often just a short little link. Worse, when the proposal is accepted, we don't have a good like "record" of accepted proposals.

The second is that it seems to be nobody's job to tag and close the issues when the FCP ends. I often go over them during compiler-team meetings. I think that #t-compiler/wg-prioritization could perhaps handle this though.

I think my proposal is that:

Now we have a record (in the form of the issue header) of the MCP.

nikomatsakis (Jun 04 2020 at 15:14, on Zulip):

This also fits with what we've been discussing in the lang team process.

simulacrum (Jun 04 2020 at 15:15, on Zulip):

I would flip the proposal to be the first thing in the issue description rather than the last thing

Santiago Pastorino (Jun 04 2020 at 15:16, on Zulip):

nikomatsakis said:

So Santiago Pastorino there are two things I don't like about MCPs "as practiced today".

The first is that we often put the design in an "out of line hackmd or gist", and the template is maybe too wordy. I find it's kind of hard to find the actual proposal in there, it's often just a short little link. Worse, when the proposal is accepted, we don't have a good like "record" of accepted proposals.

The second is that it seems to be nobody's job to tag and close the issues when the FCP ends. I often go over them during compiler-team meetings. I think that #t-compiler/wg-prioritization could perhaps handle this though.

I think my proposal is that:

Now we have a record (in the form of the issue header) of the MCP.

I think we can definitely do that work

nikomatsakis (Jun 04 2020 at 15:16, on Zulip):

simulacrum said:

I would flip the proposal to be the first thing in the issue description rather than the last thing

yeah I considered that -- seems ok

nikomatsakis (Jun 04 2020 at 15:16, on Zulip):

the main reason I was thinking "end" is that it gives us a chance to explain what the issue is first, but that's not necessary I guess

Santiago Pastorino (Jun 04 2020 at 15:17, on Zulip):

couldn't we just have a template that is more like a link to some explanations/rules/etc and then the proposal per se?

nikomatsakis (Jun 04 2020 at 15:21, on Zulip):

I think the template would be like

Proposal

(insert your proposal here)

Process

This is a Major Change Proposal. It is used to propose a change to the compiler. It works like this:

Note that this issue is meant to be used for status updates, not discussion. Discussion takes place on Zulip. You can read more about Major Change Proposals on forge.

nikomatsakis (Jun 04 2020 at 15:21, on Zulip):

short and simple enough

Santiago Pastorino (Jun 04 2020 at 15:59, on Zulip):

:+1:

Santiago Pastorino (Jun 04 2020 at 16:00, on Zulip):

nikomatsakis said:

The second is that it seems to be nobody's job to tag and close the issues when the FCP ends. I often go over them during compiler-team meetings. I think that #t-compiler/wg-prioritization could perhaps handle this though.

can you explain also the different states and when things change from one to another?

Santiago Pastorino (Jun 04 2020 at 16:01, on Zulip):

I don't remember if just one person seconding a proposal is what we need to call it accepted and then tag and close or what's the thing we need to do

Santiago Pastorino (Jun 04 2020 at 16:02, on Zulip):

in any case, we have our prioritization procedure, I'm going to write something about this there and maybe once done you can take a look at it to confirm if it's right or not

Santiago Pastorino (Jun 05 2020 at 17:35, on Zulip):

so about accepted MCPs, what I can do after weekly meetings is ... go over the list that are seconded and if we're over 10 days just tag them with major-change-accepted and close them

Santiago Pastorino (Jun 05 2020 at 17:36, on Zulip):

also, there are a lot of accepted and closed issues with no major-change-accepted tag

Santiago Pastorino (Jun 05 2020 at 17:36, on Zulip):

I can also go over them and tag now

Santiago Pastorino (Jun 05 2020 at 17:36, on Zulip):

@nikomatsakis please confirm that this is correct

Santiago Pastorino (Jun 05 2020 at 17:37, on Zulip):

actually, maybe we should more clearly state that the waiting period is 10 days + remaining days for next weekly meeting, so it may even be 16 days

Santiago Pastorino (Jun 05 2020 at 17:38, on Zulip):

it may be good idea to make this clear, so people know what to expect :)

nikomatsakis (Jun 05 2020 at 19:11, on Zulip):

I think the waiting period is 10 days

nikomatsakis (Jun 05 2020 at 19:11, on Zulip):

somewhere in those 10 days, a meeting occurs

nikomatsakis (Jun 05 2020 at 19:12, on Zulip):

I also think it's fine if we close them on some schedule so sometimes they stay "open" longer

nikomatsakis (Jun 05 2020 at 19:12, on Zulip):

ideally, we'd post a comment after 10 days

nikomatsakis (Jun 05 2020 at 19:12, on Zulip):

(automatically)

nikomatsakis (Jun 05 2020 at 19:12, on Zulip):

although the question is:

nikomatsakis (Jun 05 2020 at 19:12, on Zulip):

it's possible that during those 10 days someone may have raised a concern

nikomatsakis (Jun 05 2020 at 19:12, on Zulip):

we don't have automation around that right now

nikomatsakis (Jun 05 2020 at 19:12, on Zulip):

it's never happened yet :)

nikomatsakis (Jun 05 2020 at 19:12, on Zulip):

and it could sort of happen on zulip

nikomatsakis (Jun 05 2020 at 19:13, on Zulip):

I guess that we could/should clarify the "process" there if we want

nikomatsakis (Jun 05 2020 at 19:13, on Zulip):

we could also post a comment saying something like

The 10 days have expired. Assuming no objections were raised here or on Zulip, the MCP is accepted, and someone will come along and close the issue in a few days.

Santiago Pastorino (Jun 05 2020 at 21:06, on Zulip):

:+1:

Santiago Pastorino (Jun 05 2020 at 21:06, on Zulip):

nikomatsakis said:

somewhere in those 10 days, a meeting occurs

what I meant is that, given that we will close after a weekly meeting it may end being more than 10 days

Santiago Pastorino (Jun 05 2020 at 21:06, on Zulip):

but yeah, ideally we would be doing that in 10 days or we would be automating it

Santiago Pastorino (Jun 05 2020 at 21:07, on Zulip):

anyway, should I label closed accepted proposals?

nikomatsakis (Jun 08 2020 at 20:13, on Zulip):

yes please

nikomatsakis (Jun 08 2020 at 20:14, on Zulip):

I forgot we had a label!

nikomatsakis (Jun 08 2020 at 20:24, on Zulip):

Posted a new template: https://github.com/rust-lang/compiler-team/pull/305

nikomatsakis (Jun 08 2020 at 20:24, on Zulip):

@Santiago Pastorino, @simulacrum, take a look maybe?

Santiago Pastorino (Jun 08 2020 at 20:38, on Zulip):

:+1:, I've seen you have done some reviews already but looks good

nikomatsakis (Jun 08 2020 at 20:39, on Zulip):

thanks for adding the "accepted" labels btw

Santiago Pastorino (Jun 08 2020 at 20:45, on Zulip):

btw unsure about T-compiler labels there, there are some with T-compiler and others without it

Santiago Pastorino (Jun 08 2020 at 20:45, on Zulip):

but also I was wondering isn't everything T-compiler there?

Santiago Pastorino (Jun 08 2020 at 20:45, on Zulip):

maybe it's irrelevant, unsure :)

nikomatsakis (Jun 08 2020 at 20:53, on Zulip):

yeah we should just add it in the template

nikomatsakis (Jun 08 2020 at 20:53, on Zulip):

the only reason to have it is

nikomatsakis (Jun 08 2020 at 20:53, on Zulip):

if you do rfcbot fcp merge

nikomatsakis (Jun 08 2020 at 20:53, on Zulip):

it is needed

pnkfelix (Jun 11 2020 at 13:53, on Zulip):

Hey @Santiago Pastorino , on the subject of the MCP procedure: How hard would it be to automatically add a link to the associated Zulip thread from the MCP issue? Right now the template references "A Zulip topic in the stream #t-compiler/major changes will be created for this issue", but there's no standard place for a link to that Zulip topic, let alone a bot-generated link

Santiago Pastorino (Jun 11 2020 at 13:55, on Zulip):

I'm not sure what's the bot that does that exactly but I guess it should be easy to do it

Santiago Pastorino (Jun 11 2020 at 13:56, on Zulip):

I guess simulacrum implemented this ?

simulacrum (Jun 11 2020 at 13:57, on Zulip):

generating the link proved pretty hard when I was initially doing this

simulacrum (Jun 11 2020 at 13:57, on Zulip):

I may have been missing something though

Santiago Pastorino (Jun 11 2020 at 13:58, on Zulip):

ohh really?

Santiago Pastorino (Jun 11 2020 at 13:58, on Zulip):

I'd have thought that it was gonna be easy

Vadim Petrochenkov (Jul 01 2020 at 12:25, on Zulip):

Should MCPs be created for compiler team RFCs?

Vadim Petrochenkov (Jul 01 2020 at 12:26, on Zulip):

I'm not sure that many compiler team members are subscribed to the rfcs repo to receive the relevant notifications automatically.

Vadim Petrochenkov (Jul 01 2020 at 12:27, on Zulip):

And regular "ping @rust-lang/compiler" is such an old way to ping people that most of them developed immunity to it long ago.

nikomatsakis (Jul 01 2020 at 14:21, on Zulip):

I feel like it doesn't make sense to make both, but I think you could easily open a topic in the #t-compiler/major changes stream if you wanted to help draw attention to it

Vadim Petrochenkov (Sep 03 2020 at 20:37, on Zulip):

Ok, it looks like RFC as a process is effectively retired for the compiler team now.
All T-Compiler PRs to the RFC repo are stalled for months - https://github.com/rust-lang/rfcs/pulls?q=is%3Apr+is%3Aopen+label%3AT-compiler+sort%3Aupdated-desc (except for the aarch64 tier, which is mostly about infra).

Vadim Petrochenkov (Sep 03 2020 at 20:38, on Zulip):

I'll submit the one that I care about ("Linking modifiers for native libraries") as an MCP instead.

nikomatsakis (Sep 04 2020 at 15:10, on Zulip):

@Vadim Petrochenkov that...is probably correct

nikomatsakis (Sep 04 2020 at 15:10, on Zulip):

I didn't even realize those RFCs were there -- or at least I forgot -- I'm sorry

nikomatsakis (Sep 04 2020 at 15:10, on Zulip):

maybe we should add them to the triage meeting...

Last update: Sep 28 2020 at 15:15UTC