Stream: wg-traits

Topic: design meeting 2020.01.06


Jack Huey (Jan 06 2020 at 18:52, on Zulip):

Hi @WG-traits A bit early, but I figured I would start this so we can maybe start listing possible topics for today?

nikomatsakis (Jan 06 2020 at 18:53, on Zulip):

I was just catching up on things

nikomatsakis (Jan 06 2020 at 18:53, on Zulip):

I would say:

nikomatsakis (Jan 06 2020 at 18:53, on Zulip):
nikomatsakis (Jan 06 2020 at 18:53, on Zulip):
Jack Huey (Jan 06 2020 at 18:53, on Zulip):

ha, I'm also writing a few things

nikomatsakis (Jan 06 2020 at 18:54, on Zulip):
nikomatsakis (Jan 06 2020 at 18:54, on Zulip):
nikomatsakis (Jan 06 2020 at 18:54, on Zulip):
Jack Huey (Jan 06 2020 at 18:55, on Zulip):

Those all sound good

Jack Huey (Jan 06 2020 at 18:55, on Zulip):

In addition I would say:

detrumi (Jan 06 2020 at 18:55, on Zulip):

So many things to talk about :slight_smile:

Jack Huey (Jan 06 2020 at 18:56, on Zulip):
nikomatsakis (Jan 06 2020 at 18:56, on Zulip):

haven't caught up with that thread yet

Jack Huey (Jan 06 2020 at 18:56, on Zulip):
nikomatsakis (Jan 06 2020 at 18:57, on Zulip):
Jack Huey (Jan 06 2020 at 19:00, on Zulip):

So, how about we tentatively shoot for the following:

detrumi (Jan 06 2020 at 19:00, on Zulip):

Ah, was gonna suggest to start with one of the higher-level things, to decide on the next steps. So either tracing, general roadmap or chalkir/ty plans

David Barsky (Jan 06 2020 at 19:00, on Zulip):

(sounds good to me; hello!)

Jack Huey (Jan 06 2020 at 19:01, on Zulip):

I would like to talk about tracing a bit (@David Barsky did you still want to open a PR for that?). So maybe we can add an explicit 5-10 minutes for that

Jack Huey (Jan 06 2020 at 19:02, on Zulip):

Anyways, it's 2:01 so let's start?

nikomatsakis (Jan 06 2020 at 19:02, on Zulip):

I think the general roadmap should maybe come first, but otherwise seems reasonable

Jack Huey (Jan 06 2020 at 19:02, on Zulip):

Sure, we can start with that

nikomatsakis (Jan 06 2020 at 19:02, on Zulip):

In terms of general "sync up"...

Jack Huey (Jan 06 2020 at 19:02, on Zulip):

Do you want to summarize your convo with @pnkfelix ?

David Barsky (Jan 06 2020 at 19:03, on Zulip):

@Jack Huey i should submit the PR, but i've gotten a bit side-tracked with fixing some bugs, writing docs, and overall, taking a break. I also started mimicking the hierarchal profiling data that rust-analyzer uses

Jack Huey (Jan 06 2020 at 19:03, on Zulip):

Jack Huey i should submit the PR, but i've gotten a bit side-tracked with fixing some bugs, writing docs, and overall, taking a break. I also started mimicking the hierarchal profiling data that rust-analyzer uses

No problem :) I'm pretty excited about it

nikomatsakis (Jan 06 2020 at 19:04, on Zulip):

Sure. The TL;DR is that we were discussing the question of how much to focus on integrating chalk vs extending rustc, and also looking a bit into what the challenges are around integrating chalk and if we can do it a bit more "Step by step"

detrumi (Jan 06 2020 at 19:04, on Zulip):

Step by step?

nikomatsakis (Jan 06 2020 at 19:05, on Zulip):

I think the outcome was that I was thinking that (a) I think I sort of see the path to extend rustc to handle more of GATs natively, and hopefully to handle lazy norm better, which before I wasn't sure if I saw such a path, and (b) but I still feel like I want to pursue the overall chalk model

nikomatsakis (Jan 06 2020 at 19:05, on Zulip):

by "step by step" what I meant really was

nikomatsakis (Jan 06 2020 at 19:05, on Zulip):

are there ways to make chalk behave "more like" the existing trait solver so we can integrate more easily. I'm not sure about that, I think there might be some room to do so in terms of exploring different kinds of solvers

Jack Huey (Jan 06 2020 at 19:06, on Zulip):

Can you elaborate on different kinds of solvers?

nikomatsakis (Jan 06 2020 at 19:07, on Zulip):

I think that as far as chalk integration goes, we should really be looking at integrating chalk-solve into rustc (with a -Z flag) as a next goal, although I'm not sure how best to balance with "responding to problems rust-analyzer finds"

Jack Huey (Jan 06 2020 at 19:08, on Zulip):

Which is a few at this point

nikomatsakis (Jan 06 2020 at 19:08, on Zulip):

Can you elaborate on different kinds of solvers?

there are different ways to structure the solver. Chalk used to have a different solver that worked more like rustc, in that it didn't try to enumerate individual solutions -- it looks always for unique answers. I'm trying to remember what some of the problems were that we had with it, I do remember it sometimes was failing to find solutions.

nikomatsakis (Jan 06 2020 at 19:08, on Zulip):

Yeah, it seems like they're [rust analyzer problems] starting to accumulate

Jack Huey (Jan 06 2020 at 19:09, on Zulip):

(Good news is that I can reproduce a couple with existing tests by just setting max_size to 3)

nikomatsakis (Jan 06 2020 at 19:09, on Zulip):

Yeah I'm also quite curious about the infinite loops that arise with higher max-size values

Jack Huey (Jan 06 2020 at 19:10, on Zulip):

I think we need some much more complex tests in chalk to start to see those

detrumi (Jan 06 2020 at 19:10, on Zulip):

Integrating with rustc should be easier after we've fixed any large problems from the currently found bugs

Jack Huey (Jan 06 2020 at 19:10, on Zulip):

I don't any now pass a max_size of like 6

nikomatsakis (Jan 06 2020 at 19:10, on Zulip):

I'm not sure if that's true; I would expect rustc integration to be sort of orthogonal

nikomatsakis (Jan 06 2020 at 19:11, on Zulip):

I think the immediate steps there are going to be finishing up the work on mapping types

nikomatsakis (Jan 06 2020 at 19:11, on Zulip):

and then trying to implement the chalk-rust-ir callbacks

Jack Huey (Jan 06 2020 at 19:11, on Zulip):

I guess that kind of leads into the chalk-ir update?

nikomatsakis (Jan 06 2020 at 19:11, on Zulip):

yeah, they're definitely connected

nikomatsakis (Jan 06 2020 at 19:12, on Zulip):

I guess my main take-away from the conversation with felix was that there is a reasonable path to both improve rustc and keep working on chalk

Jack Huey (Jan 06 2020 at 19:12, on Zulip):

is that sort of blocked on compiler team meeting?

nikomatsakis (Jan 06 2020 at 19:12, on Zulip):

I need to write-up the plans but some of them are already in progress

nikomatsakis (Jan 06 2020 at 19:12, on Zulip):

is that sort of blocked on compiler team meeting?

I don't think so

nikomatsakis (Jan 06 2020 at 19:12, on Zulip):

I need to write-up the plans but some of them are already in progress

that said, the people in this meeting I thnk are mostly folks who've been hacking on chalk so let's focus on that part of things for now

nikomatsakis (Jan 06 2020 at 19:13, on Zulip):

so as far as chalk-ir goes, I mean, I think there is some question as to what the best way is to proceed-- I've kind of been piecemeal making refactorings, but probably the right step is to finish sketching out the end-state and create an issue to help others make some of those changes?

Jack Huey (Jan 06 2020 at 19:13, on Zulip):

So, for that, the next steps are to basically write PRs to change chalk-ir to match the proposal? (which you've been doing I think)

Jack Huey (Jan 06 2020 at 19:13, on Zulip):

ninjad

nikomatsakis (Jan 06 2020 at 19:13, on Zulip):

yeah, but there are a few big changes left :)

nikomatsakis (Jan 06 2020 at 19:14, on Zulip):

in particular

nikomatsakis (Jan 06 2020 at 19:14, on Zulip):

I think that if we are going to support interning

nikomatsakis (Jan 06 2020 at 19:14, on Zulip):

we need to be passing around a tcx: &TypeFamily or something; maybe it should be named something else of course

nikomatsakis (Jan 06 2020 at 19:14, on Zulip):

and the various intern_ty functions need to become &self methods

nikomatsakis (Jan 06 2020 at 19:15, on Zulip):

this will be an annoying refactoring :)

nikomatsakis (Jan 06 2020 at 19:15, on Zulip):

maybe it should be named something else of course

(probably a name like interner: TypeInterner is the most obvious, side note)

Jack Huey (Jan 06 2020 at 19:16, on Zulip):

I really need to read through the current chalk-ir code and get a more general sense of everything with some of the newer refactors

nikomatsakis (Jan 06 2020 at 19:16, on Zulip):

the other question is whether the lookup methods (e.g., the one used by ty.data()) should be &self; this would permit interning that is creating integers and be more general, but also more annoying. I would think we would first just make the intern calls into methods

detrumi (Jan 06 2020 at 19:17, on Zulip):

Not sure what this would look like

Jack Huey (Jan 06 2020 at 19:18, on Zulip):

(a bit of a off-topic/side thought, but this reminds me a bit of Pyo3's Py/PyObject system)

nikomatsakis (Jan 06 2020 at 19:18, on Zulip):

so I guess the remaining steps would be:

at that point I think we could take a first stab at modifying rustc to do the mapping

nikomatsakis (Jan 06 2020 at 19:18, on Zulip):

actually we could start doing it before the third point

nikomatsakis (Jan 06 2020 at 19:18, on Zulip):

since we can just panic! on missing cases

nikomatsakis (Jan 06 2020 at 19:18, on Zulip):

it'll probably help us identifying other problems

Jack Huey (Jan 06 2020 at 19:18, on Zulip):

Sounds good to me

nikomatsakis (Jan 06 2020 at 19:19, on Zulip):

to finish the rustc integration, then, we'd also have to implement the RustIrDatabase trait, which provides things like "information about an impl"

nikomatsakis (Jan 06 2020 at 19:19, on Zulip):

it all feels eminently do-able to me but will require some sustained attention I guess

Jack Huey (Jan 06 2020 at 19:19, on Zulip):

I left a couple comments on your "intern substitutions" pr, but overall looks decent

Jack Huey (Jan 06 2020 at 19:20, on Zulip):

Ok, all this sounds good to me

Jack Huey (Jan 06 2020 at 19:20, on Zulip):

Move on to general sync-up?

nikomatsakis (Jan 06 2020 at 19:21, on Zulip):

I left a couple comments on your "intern substitutions" pr, but overall looks decent

btw I just responded to your substantive question

nikomatsakis (Jan 06 2020 at 19:21, on Zulip):

one question before we move on

nikomatsakis (Jan 06 2020 at 19:21, on Zulip):

well I guess the question is: it seems clear that we can make "individual steps" for chalk-rustc integration

nikomatsakis (Jan 06 2020 at 19:21, on Zulip):

I think I can do the same for the "extend rustc to support gats etc" plan I was hinting at

nikomatsakis (Jan 06 2020 at 19:22, on Zulip):

but I think the question is (A) how much to balance the rust-analyzer and optimization work

nikomatsakis (Jan 06 2020 at 19:22, on Zulip):

and (B) who will do the work and how to help out best there

nikomatsakis (Jan 06 2020 at 19:22, on Zulip):

I'm specifically pondering if I can make more of a commitment to individual people

nikomatsakis (Jan 06 2020 at 19:22, on Zulip):

e.g., we could schedule like a regular meeting time

Jack Huey (Jan 06 2020 at 19:22, on Zulip):

So, last meeting I put up my TODO for chalk

nikomatsakis (Jan 06 2020 at 19:23, on Zulip):

I'm specifically pondering if I can make more of a commitment to individual people

(vs the more general "calls for help" that we sometimes do)

nikomatsakis (Jan 06 2020 at 19:23, on Zulip):

So, last meeting I put up my TODO for chalk

link?

Jack Huey (Jan 06 2020 at 19:24, on Zulip):

https://rust-lang.zulipchat.com/#narrow/stream/144729-wg-traits/topic/design.20meeting.202019.2E12.2E30/near/184494512

Jack Huey (Jan 06 2020 at 19:25, on Zulip):

Which I think mostly still is correct for me

nikomatsakis (Jan 06 2020 at 19:25, on Zulip):

my impression is that @Jack Huey is having a blast doing the "tune chalk" side of things :)

nikomatsakis (Jan 06 2020 at 19:25, on Zulip):

(vs say rustc-integration)

Jack Huey (Jan 06 2020 at 19:25, on Zulip):

But also after that first point is to at least investigate the bugs from rust-analyzer and see if they have "easy" fixes vs "design" fixes

nikomatsakis (Jan 06 2020 at 19:26, on Zulip):

I do think that makes a lot of sense

Jack Huey (Jan 06 2020 at 19:26, on Zulip):

I am quite enjoying chalk side of things

nikomatsakis (Jan 06 2020 at 19:26, on Zulip):

in particular my biggest fear for chalk, well, there are maybe 2

nikomatsakis (Jan 06 2020 at 19:26, on Zulip):

the first one is that the existing solver won't work out because it tries too hard to pursue individual options; I think maybe I shouldn't be as afraid about that, because part of the whole point of chalk was to have the freedom to change the solver strategy

nikomatsakis (Jan 06 2020 at 19:27, on Zulip):

but the best way to assess that fear is to look at the problems we're having, as you said

detrumi (Jan 06 2020 at 19:27, on Zulip):

The main danger there is that we make things too complicated

Jack Huey (Jan 06 2020 at 19:27, on Zulip):

I think several of these issues really only require changes in engine

Jack Huey (Jan 06 2020 at 19:28, on Zulip):

but don't quote me on that

nikomatsakis (Jan 06 2020 at 19:28, on Zulip):

I feel pretty confident this is the case

nikomatsakis (Jan 06 2020 at 19:28, on Zulip):

I'm basically not sure if the engine design is correct

nikomatsakis (Jan 06 2020 at 19:28, on Zulip):

the lowering part I think is in the correct area--though there is one big work item there (caching)

nikomatsakis (Jan 06 2020 at 19:28, on Zulip):

the lowering part seems also to be the most valuable, in that it's the first real definition of what Rust's trait system means :P

Jack Huey (Jan 06 2020 at 19:29, on Zulip):

what was your second biggest fear?

nikomatsakis (Jan 06 2020 at 19:29, on Zulip):

(caching)

and integration with incremental

nikomatsakis (Jan 06 2020 at 19:30, on Zulip):

what was your second biggest fear?

I was wondering that myself :) maybe I only have 1. I mean I think i'm just afraid of the 'long tail' of diagnostics and other work, and afraid that pursuing it will take a long time where there is a shorter path in rustc to unlock some stuff. Which is a bit why I'm interested in putting in more time this year and trying to pursue both quite seriously.

nikomatsakis (Jan 06 2020 at 19:31, on Zulip):

to elaborate a bit on the caching/incremental thing: right now, the engine asks for "what are the program clauses to solve goal X" and I think each time it does so we go and re-invoke all the same methods do all the lowering again and again

Jack Huey (Jan 06 2020 at 19:31, on Zulip):

I think you're correct with that

nikomatsakis (Jan 06 2020 at 19:31, on Zulip):

we wrote that code kind of "naively" just to stand something up, but I'm pretty sure that is a big part of some of the perf problems we're seeing -- based on that profile we were discussing, for example

Jack Huey (Jan 06 2020 at 19:32, on Zulip):

that makes sense

nikomatsakis (Jan 06 2020 at 19:32, on Zulip):

I'm not exactly sure how to do better :) I've not really tried to think about it. It seems like it should be eminently possible.

nikomatsakis (Jan 06 2020 at 19:32, on Zulip):

(as a first step, we could cache the rules for the environment, which rarely changes)

nikomatsakis (Jan 06 2020 at 19:33, on Zulip):

we have to be a bit careful but

nikomatsakis (Jan 06 2020 at 19:33, on Zulip):

the other thing is that right now, when we go to create program clauses,

nikomatsakis (Jan 06 2020 at 19:33, on Zulip):

we always recreate them, re-querying the database

nikomatsakis (Jan 06 2020 at 19:34, on Zulip):

anyway let's table the deatils I guess, I imagine we'd want to go and gather some data

Jack Huey (Jan 06 2020 at 19:35, on Zulip):

Do we want to do a quick sync-up? Then jump to tracing?

nikomatsakis (Jan 06 2020 at 19:35, on Zulip):

Sorry, we were going to move on to general sync up .. maybe we can summarize what we said? I think it was:

Three high-level goals:

Jack Huey (Jan 06 2020 at 19:36, on Zulip):

sounds good to me

Jack Huey (Jan 06 2020 at 19:37, on Zulip):

I can definitely write out some goals for chalk tuning

nikomatsakis (Jan 06 2020 at 19:37, on Zulip):

As for a quick sync-up:

I did a bit of work towards chalk-ir. I want to merge the book PR and try to finish it up. I am also curious where people think are the best place to "write out the steps" and try to collect all the work that's going on. (Tracking issues? The wg-trait repository? I'm curious maybe to try Github Projects on wg-traits combined with issues from other repositories)

nikomatsakis (Jan 06 2020 at 19:37, on Zulip):

(also curious if people think ti'd be better to merge book PR with chalk-ir proposal and then write more, or try to finish it in PR form)

Jack Huey (Jan 06 2020 at 19:38, on Zulip):

So, I think merging the book PR proposal as it is now is good

Jack Huey (Jan 06 2020 at 19:38, on Zulip):

As far as writing out the steps:

Jack Huey (Jan 06 2020 at 19:39, on Zulip):

Probably a tracking issue on rust-lang/chalk for refactorings. Then if any are big enough that could go in parts, maybe split/add more

Jack Huey (Jan 06 2020 at 19:39, on Zulip):

I think a wg-traits issue would be good for rustc integration

nikomatsakis (Jan 06 2020 at 19:40, on Zulip):

I think I am going to try github projects, why not, although I don't know how useful it'll be

nikomatsakis (Jan 06 2020 at 19:41, on Zulip):

I can't quite tell what the columns should be, in particular :) but we'll see

Jack Huey (Jan 06 2020 at 19:42, on Zulip):

As for my sync-up:

detrumi (Jan 06 2020 at 19:42, on Zulip):

rust-clippy uses 'todo' and 'done' on their projects page, that seems to work

nikomatsakis (Jan 06 2020 at 19:43, on Zulip):

created a GH project for rustc-chalk integration, will experiment more with it later though :)

nikomatsakis (Jan 06 2020 at 19:44, on Zulip):

(On the note of associated types stuff, I still want to explore the "split syntactic and semantic equality" and then experiment a bit with having chalk solve for region constraints -- that's definitely a design question, and I think a more general one.)

Jack Huey (Jan 06 2020 at 19:44, on Zulip):

I've sort of been trying to implement what I said in the other topic: instead of marking an exclause as ambiguous, I just add a CannotProve delayed subgoal

nikomatsakis (Jan 06 2020 at 19:44, on Zulip):

hmm

nikomatsakis (Jan 06 2020 at 19:44, on Zulip):

yeah I didn't have time to process

nikomatsakis (Jan 06 2020 at 19:44, on Zulip):

but I can sort of imagine why that might be useful

Jack Huey (Jan 06 2020 at 19:44, on Zulip):

yeah, I figured if you had read through and thought about it, you would have replied :)

Jack Huey (Jan 06 2020 at 19:45, on Zulip):

(I think really, the goal would be to use specific goals for "ambiguity" instead of CannotProve)

Jack Huey (Jan 06 2020 at 19:45, on Zulip):

So you could, theoretically "pick up" from a truncation

nikomatsakis (Jan 06 2020 at 19:46, on Zulip):

should we discuss tracing a bit here at the end?

Jack Huey (Jan 06 2020 at 19:46, on Zulip):

yes, I was just gonna say that

David Barsky (Jan 06 2020 at 19:46, on Zulip):

guess so?

nikomatsakis (Jan 06 2020 at 19:46, on Zulip):

Not sure if @David Barsky is still around

David Barsky (Jan 06 2020 at 19:47, on Zulip):

heirarchal.png i was working on a hierarchal fmt for tracing to mimic what rust-analyzer has (and because tree representations are handy)

David Barsky (Jan 06 2020 at 19:48, on Zulip):

to interpret the data: each bolded green word conn, server, sloggish-example is a span, correspnding to a unit of of work

Jack Huey (Jan 06 2020 at 19:48, on Zulip):

I think it will actually be pretty straightforward to convert the CHALK_DEBUG calls to tracing

nikomatsakis (Jan 06 2020 at 19:48, on Zulip):

I should think so-- but I think we could do better

Jack Huey (Jan 06 2020 at 19:48, on Zulip):

since there are debug_heading and info_heading which are sort of spans

nikomatsakis (Jan 06 2020 at 19:48, on Zulip):

e.g., I'd prefer if we assigned each strand a unique id

David Barsky (Jan 06 2020 at 19:48, on Zulip):

the log-like messages within in are log-style messages

nikomatsakis (Jan 06 2020 at 19:48, on Zulip):

and we use that to tag things

Jack Huey (Jan 06 2020 at 19:48, on Zulip):

Yes, that would be really neat

nikomatsakis (Jan 06 2020 at 19:48, on Zulip):

similarly we can tag events with their table, so you can view events by table

nikomatsakis (Jan 06 2020 at 19:49, on Zulip):

I'm not quite sure what all the ways are that you would want to view it I guess

Jack Huey (Jan 06 2020 at 19:49, on Zulip):

This honestly makes me so excited

David Barsky (Jan 06 2020 at 19:49, on Zulip):

i'd like to have this print out _durations_, rather than instants. this requires some small changes in terms of how this data is outputted

nikomatsakis (Jan 06 2020 at 19:49, on Zulip):

I feel like @David Barsky and @Jack Huey doing this together would be a power team hint hint ;)

Jack Huey (Jan 06 2020 at 19:49, on Zulip):

Yes

David Barsky (Jan 06 2020 at 19:49, on Zulip):

(e.g., not when events come in—they need to be buffered and flushed on a span's exit)

David Barsky (Jan 06 2020 at 19:50, on Zulip):

yep, those are all good use-cases

Jack Huey (Jan 06 2020 at 19:50, on Zulip):

@David Barsky if you make an initial PR, I can definitely add more

nikomatsakis (Jan 06 2020 at 19:50, on Zulip):

one challenge, but presumably it can be handled,

nikomatsakis (Jan 06 2020 at 19:50, on Zulip):

is to have a nice way to deal with "strand A requests a result from table B"

nikomatsakis (Jan 06 2020 at 19:51, on Zulip):

i.e., to be able to trace all the work done on behalf of some parent strand?

David Barsky (Jan 06 2020 at 19:51, on Zulip):

@Jack Huey lemme finish this formatter up first + get that published, but i've rewritten the getting started guide: https://github.com/tokio-rs/tracing#usage

nikomatsakis (Jan 06 2020 at 19:51, on Zulip):

anyway it seems like exactly the sort of problem that tracing was built to solve though

David Barsky (Jan 06 2020 at 19:51, on Zulip):

hmm

Jack Huey (Jan 06 2020 at 19:52, on Zulip):

Should make a topic for this so we can talk about it async

David Barsky (Jan 06 2020 at 19:52, on Zulip):

i _think_ you can you can treat that as a follows_from or parent/child relationship in tracing

David Barsky (Jan 06 2020 at 19:52, on Zulip):

yep

Jack Huey (Jan 06 2020 at 19:52, on Zulip):

I don't think we had an issue about it yet

David Barsky (Jan 06 2020 at 19:53, on Zulip):

but yes, if your call graph is (mostly) a DAG, tracing is a good fit

Jack Huey (Jan 06 2020 at 19:53, on Zulip):

Neat

Jack Huey (Jan 06 2020 at 19:54, on Zulip):

Ok, anything else to mention in the remaining 6 minutes?

Jack Huey (Jan 06 2020 at 19:54, on Zulip):

I think we actually covered everything we said we wanted to

Jack Huey (Jan 06 2020 at 19:55, on Zulip):

which is somewhat surprising

nikomatsakis (Jan 06 2020 at 19:55, on Zulip):

Heh go team :heart:

Jack Huey (Jan 06 2020 at 19:55, on Zulip):

Ok, so I'll just make a quick summary of the meeting then:

Jack Huey (Jan 06 2020 at 19:56, on Zulip):

(and things to do next)

nikomatsakis (Jan 06 2020 at 20:00, on Zulip):

(are you posting that here?)

Jack Huey (Jan 06 2020 at 20:01, on Zulip):
Jack Huey (Jan 06 2020 at 20:01, on Zulip):

Yes :)

Jack Huey (Jan 06 2020 at 20:01, on Zulip):

Did I miss anything

nikomatsakis (Jan 06 2020 at 20:01, on Zulip):

Sounds great, also, we're going to create a projects on wg-trait to try to get organized

nikomatsakis (Jan 06 2020 at 20:02, on Zulip):

I think it'd be nice to try and create a plan of sorts for "the next quarter" or so -- maybe 6 weeks is a good choice -- and write a blog post about it

nikomatsakis (Jan 06 2020 at 20:02, on Zulip):

seems like we're close

Jack Huey (Jan 06 2020 at 20:02, on Zulip):

Sounds great, also, we're going to create a projects on wg-trait to try to get organized

thanks, added

Jack Huey (Jan 06 2020 at 20:03, on Zulip):

I think a blog post would be cool

Jack Huey (Jan 06 2020 at 20:03, on Zulip):

I would say maybe 10 weeks?

nikomatsakis (Jan 06 2020 at 20:03, on Zulip):

depends on the scale of the goals

Jack Huey (Jan 06 2020 at 20:03, on Zulip):

true

nikomatsakis (Jan 06 2020 at 20:04, on Zulip):

I only chose 6 weeks because it's rust release cadence anyway, doesn't matter ;)

Last update: Jun 07 2020 at 10:35UTC