Stream: t-compiler/wg-nll

Topic: rfc/spec updates


Jake Goulding (Sep 14 2018 at 14:06, on Zulip):

I think we ought to revisit the NLL RFC

I'm a big fan of this kind of thing happening. I think just a new section at the end of the RFC with "differences between the RFC and the real world" would be sufficient. The reference should be the ultimate authority, but it is rarely touched. I think that it's too scary for most people to edit (once it's in there, it's THE FINAL WORD, oh noes), so people read the RFC to figure out things.

Jake Goulding (Sep 14 2018 at 14:08, on Zulip):

And of course, this stuff also gets "documented" in GitHub, Stack Overflow, URLO/IRLO, Reddit, etc., but that's not great either.

pnkfelix (Sep 14 2018 at 14:15, on Zulip):

(context: https://rust-lang.zulipchat.com/#narrow/stream/122657-wg-nll/subject/making.20plans/near/133735334 )

Jake Goulding (Sep 14 2018 at 14:16, on Zulip):

Ah yes, sorry. I assumed that quoting would do that for me (I'm catching up on ~2 weeks of backlog :wink: )

pnkfelix (Sep 14 2018 at 14:19, on Zulip):

unfortunately that's one place where Zulip misses out (the fact that the "quote" command does not actually hyperlink when it could)

pnkfelix (Sep 14 2018 at 14:19, on Zulip):

/me wonders whether this could be worth filing an issue in Zulip's github

nikomatsakis (Sep 14 2018 at 14:42, on Zulip):

I think the problem with the reference is that we've historically not "vetted" its contents

nikomatsakis (Sep 14 2018 at 14:42, on Zulip):

I'm sure it's accurate etc

nikomatsakis (Sep 14 2018 at 14:43, on Zulip):

I would like for 2019 to start an effort to change that

nikomatsakis (Sep 14 2018 at 14:43, on Zulip):

I know that @alercah and @havvy have done awesome work here anyway

nikomatsakis (Sep 14 2018 at 14:43, on Zulip):

tl;dr I think we should put this text into the reference, but we can also have a summary+link in the RFC

nikomatsakis (Sep 14 2018 at 14:43, on Zulip):

well, I'm not sure if that's my total feels

nikomatsakis (Sep 14 2018 at 14:43, on Zulip):

I had some thoughts about the best process here ...

nikomatsakis (Sep 14 2018 at 14:44, on Zulip):

...I think I may have talked about them in that post on a propsoed new RFC process

Jake Goulding (Sep 14 2018 at 15:16, on Zulip):

I'm sure it's accurate etc

DYM "I'm not sure"?

Jake Goulding (Sep 14 2018 at 15:17, on Zulip):

I'd also think that the reference could stand to have some styling thing that outlines a section as non-normative / non-binding / non-vetted whatever

nikomatsakis (Sep 14 2018 at 15:18, on Zulip):

DYM "I'm not sure"?

what I really meant was "I'm sure it's 99% fine"

nikomatsakis (Sep 14 2018 at 15:18, on Zulip):

but I also think

nikomatsakis (Sep 14 2018 at 15:18, on Zulip):

there is no such thing as a normative spec :)

nikomatsakis (Sep 14 2018 at 15:18, on Zulip):

like, spec bugs are a thing

Jake Goulding (Sep 14 2018 at 15:19, on Zulip):

certainly.

nikomatsakis (Sep 14 2018 at 15:19, on Zulip):

what I really mean is: I think there are attempts at normative specs ,but ultimately, when you find a discrepancy between a spec + an impl, it's a time to do inquiry and figure out what the story should be...

nikomatsakis (Sep 14 2018 at 15:19, on Zulip):

(of course the longer the spec has been around etc the more normative it is... it's a spectrum)

Jake Goulding (Sep 14 2018 at 15:19, on Zulip):

metadata on every sentence (word?) with a "confidence percentage" :wink:

nikomatsakis (Sep 14 2018 at 15:19, on Zulip):

tl;dr think we can just say "This is the spec. Let us know if you find a discrepancy. :)"

nikomatsakis (Sep 14 2018 at 15:20, on Zulip):

but I think we should also start and maintain an effort to write tetsts against the spec

nikomatsakis (Sep 14 2018 at 15:20, on Zulip):

and proactively find such discrepancies

nikomatsakis (Sep 14 2018 at 15:20, on Zulip):

plus fill it out in places where it is lacking

nikomatsakis (Sep 14 2018 at 15:20, on Zulip):

some things (e.g., formalizing every detail of typeck) are very hard, we should not start there

nikomatsakis (Sep 14 2018 at 15:20, on Zulip):

anyway I know that there's been awesome work done here and I am sad that I haven't really reviewied it in depth

Last update: Nov 21 2019 at 13:10UTC