.Net Foundation Design Reviews, 4 Dec 2015

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: .NET Core Design Reviews: Backlog review

Description

00:00:00 - Approved: Missing several valuable members of System.Console in contract https://github.com/dotnet/corefx/issues/4636#issuecomment-162039060
00:10:27 - Approved: Add ValueTask to corefx https://github.com/dotnet/corefx/issues/4708#issuecomment-162042013

A

That's a good awesome, Stephen kin years, I can't.

B

Can hear me awesome.

A

It just works, it's surprising what you boys.

B

Did you change your microphone? Did I change our microphone yeah.

A

B

Same when I've used so I'll headset, thingy Wow.

C

Like it's really good excellent.

B

C

B

Some of the voice, lessons I've been taking.

A

Okay, so now I just need to find a way to project what actually the project, and so that's not as easy as you think.

C

So Steve, what do you think about the process? Api's discussion.

B

I think the new api's are nice. I still have trouble with the notion of introducing a new set of API is even though you know, the current ones aren't great. They support the same scenarios and so I I struggle with the idea of introducing new ones. I, don't completely understand the whole standard, plat versus standard lib, vs court effects, distinction, and so I I, don't quite know what to make of it. But I, like the mean in nothing with you, know all other issues aside. The new shape looks nice.

A

Yeah I think that's something we need to talk about in general. Is that what we do with our competitor, because the thing in general, like I, think we said it before? If you need to think about how we can evolve, don't make war without moving the elephant every single time. You add something, um but at the same time there's a lot of contention about what it means.

A

There's a combat effort. Hopefully, I can publish something today: special threats. What we saw the threads on home ah github regarding our ATO done it interfaces and I think we need to put a bit stronger. Put it up.

A

You know foot in the door where we can actually talk about what that looks like and why we adding API is about what not adding API means and what I wrote looks like and I think once you have, that I think those discussions would hopefully be a bit easier but I think in general, I, don't know it's hard to evolve, something if you can't modify it and it's also hard to evolve, something that you can add.

A

You know alternatives right and so I think we need to find a way we can do one or the other.

A

All right, so here's what I projected can you share this? Can you see the screen Stephen uh yeah I can take the land that works can find my cursor here. It is awesome, so.

B

So scroll to the bottom rock there you go right there: okay, so uh it's pallabi there.

A

Unfortunately, no with the only people here is crystal myself in you: okay! Well, that's like that that, because we don't have many people attending, because Wes does a one night like recording yeah.

B

That's fine, so I there may be more stuff on console that we want to bring back, but these were the ones that jumped out at me and that I saw being used in a variety of places, including there was a whole discussion around. You know implementing something like more advanced cursor manipulation as part of doing standard. Read line stuff. You know you want to read a line of text and have a history and be able to go back and forth and things you've already typed and whatnot.

B

And if you look at the solutions that are out there that are in c-sharp, they all use things like set cursor position and clear and whatnot yeah and any kind of standard. You know any console application that does anything real basically needs all this stuff.

B

The one exception- and we can talk about this as buffer with buffer height, so I threw those in because one particular library was using them, but on linux, they're just going to return window with window height and unless we expose the rest of the buffer related api's, which I didn't do they're sort of meaningless other than to help existing code compile- and you probably also, I think, there's also like a buffer left in a buffer top with I'm, not positive, so you probably want all of them.

B

But anyhow, these are the ones that I I added implementations for to linux, but I didn't expose them in the contract and I didn't copy over the windows, implementation from the reference source, yet I think assuming we decided to go ahead with them. Pallabi was going to do both of those things so.

A

I think and I talked to Wes about console I, don't know 100 * day at this point, I think the general consensus that we all agreed proved is that we should check the holistic local console pretty much at everything except for the ones. We don't want to add specifically because right now its basis is the one of all I went into this guy eat, so they be either. Add this one right, I great.

B

A

I think autism.

B

C

Ahead cough so I, don't see on the oldest on this list. The.

A

Counter manipulation in there we've already edit those right there. Already there good.

C

Oh really yeah, but.

B

C

More like two different color ABI, so Alice, you change the color than you write in a different color. Yes, and then there were those where you can actually do like. You know: grab place, II graphics.

C

They are related to the cursor manipulation. Putting.

A

I mean like the way our College icicle.

B

The only color ap icing- oh.

C

Really maybe I forgot: okay, okay,.

A

C

A

That we have I think that's the only the one you want is there's a this one, specific API that that's essentially a copy of the buffer, where you can say take this thing over here and copy it here. I would like to do pray, for example, scrolling without having to in water screen, okay, but I'll call AP, I suggest you said foreground and background images right. Oh.

C

And what is their cursor left in person? Talk this.

A

Face is the location where you're right, yes, their own column, I.

C

See this tense do so. It's basically like the set cursor position except independent.

B

C

B

It's the getter for set cursor position. I, don't know why there are two separate things, but oh I.

A

Guess somebody didn't wanna do so type, no.

C

It's probably because setting it twice may be expensive and usually be one to just move the village. So.

A

I said we probably exposed to api's instead of one because nobody wants to introduce a column, roll data type right.

C

B

I always need a point or something.

A

B

Be nice if we had that because we actually have to you know if you want to get the current cursor position, we have to basically to double the work yeah.

A

Yeah I mean in general, like every time, I ride a console or app that does something meaningful on the screen and all the work on tape. You guys are really slow, like it's very hard to get a flicker, free, rendering and scrolling I'm. Sorry.

C

A

What I try this was basically like git diff, but being able to stage individual lines. So it's basically like a text editor that you can scroll and you have a current highlighted line and that those things are relatively expensive. It's not nice yeah.

B

That's it I as part of validating the set of api's on linux. I did find a few different, interesting apps. I found a space invaders in c-sharp for the console, and so I have that up and running on linux. Using these.

A

So, what's your high score.

B

That's not implemented mm, but that is nothing to do with counsel.

A

Mr. priorities, here all right, so I think you for my point of view. There's nothing wrong with adding those I think the only thing I would do at this point is like, if you already looked at the consulate guys, which ones did you not include, because the thing it would be good to see is all of console which are the ones we have netted yet and having not done it yet, because we didn't run in the app. So do we not add those, because we think that the wrong API is to add Italy.

B

This sucks I think that makes the pencil. Let's say we we could go, give the thumbs up for all of these, except for buffer buffer with buffer height, because those two are tied in with some the other AP ice and then all in addition to giving the go-ahead on those I'll, add the remaining list and whether I think we should add or not, and why. Okay.

A

Let me just reply and say.

A

Looks good, except for what did you say, buff aware.

A

And but for high, because the concern was that they're meaningless, basically undone on Linux yeah.

B

I also need to look I think that there are few other buffer relative api's and we should probably either add them all or add. None of them as we just need to go through and look.

B

The only reason I ended up, including these two is because there was some code that I was trying to compile as is, and it used them, but that was it.

A

We should let it get around um and you read all the next behavior.

C

A

Then let me actually mark this going.

A

All right done.

B

Okay, so this one we meet you, the three of us have had email conversations about this for a while um and so I implemented it I have a I didn't put up a PR to follow the PAP I process, but I have it up in a commit and I'd like to basically add a new library, decor effect, so that we don't have to put this. You know into desktop and whatnot. It can just be a separate thing that sits on top of it available everywhere.

B

Basically, it just targets done in 45 api's, so right now, I'm proposing a system, threading tasks, extensions library we could choose whatever name.

B

We want imagining that in the future, if we were to, for example, add a new like a sink reader writer lock, maybe you could go in there as well um and but right now it would basically just be the value task type and the associated a waiter for it, and then, if, if there's a proposal for Rosalind to be able to have other things in the return type position of a sink methods, of which this would be one, and so we would also add the necessary builder types to that assembly.

B

If that the fat goes ahead.

C

And into whatever, as ever, you want, it was everyone who added two. If.

B

There wasn't any concern about you know, bringing it to other implementations or whatnot I put it into system starting tasks, agree.

A

With that I mean we can still edit their right, we can still support it on desktop if it's a separate time, because they do a partial facade. If you really want to do yeah.

C

Kind of, like you know recently, we've had more discussions about this here. It's right, I really don't like how is kind of twist the architecture in this factoring of our types and libraries. Because of the you know, not that framework, tesco issue, I kind of I, almost think that you should consider doing function, facade and basically, that home have.

A

It the way with me like I, think there's nothing wrong with that. The question is Stephen like their. This seems to be people like really jumping on it, and it said the savings are huge like from what I've heard an email, its use more like this is basically.

C

A

Cases where you already have the results computed, but it doesn't make the general async a cheaper right, because you basically just rub an existing task right right.

B

Variaty, so it's very helpful in the case where you have a very hot method that usually complete synchronously completes asynchronously every once in a while. That's returning sort of arbitrary values. So it's not easy to predict the value in those cash them and in those cases it can make an absolute huge savings.

B

So it's sort of a niche thing that doesn't impact most cases, but in the few cases that it does it's huge so like, for example, we use this in one place in court effects now where it had a gigantic impact in the xml library for the async aps, but that's the only place we use it. I use it one place in the channels, library, although that place, that I use it is public and there were two alternatives: either a much more expensive implementation or a much more convoluted api.

B

That required multiple calls to do the same thing. Yeah.

C

So you see like some times in the past. We avoided experience because of the cost of the class task. Like you know, binary, reader and writer, you don't want to read a bite, they say v, you know asynchronously. Yes, do you think that the volume tasks improve the situation, not such that you can actually come in? You know a single relations or something like a binary reader.

B

It's still more expensive than the synchronous call, but yeah. It certainly moves in that direction, but.

A

What's the cost the case? Because if you basically, if let's say you, you have the binary reader right, you already have the buffer, so you just return the bite from the buffer. So you find the position in the buffer. Then you return the value task where that by to that position, what what is it I mean? You just had a very thin structure at that point right, but.

B

You've got you, then every severy time you're accessing that struck. There's a brand right. You have to check to see whether it contains a task or not so which path you want to go down, you're returning a basically a double word sized thing from the method, rather than the individual character.

B

So they're, you know they're these additional overheads that they're small, but it's not as cheap as if it was the synchronous case, stuff yeah.

A

Well, the question is that give me I think that comparison is.

A

You know you do a task or, as you do it synchronously, then the Delta may be huge. But if you do the task or survey to task and the Delta between the video task and the tasks may be much much bigger than the difference between synchronous and video task right, so they may still move it into a realm where it's at least from an API for doing.

B

It yeah and my only hesitancy this type and it doesn't. I absolutely believe we should add it. My only hesitancy with it is that people are going to miss.

A

You can actually.

B

Just use it yes in the case where you're usually completing asynchronously, it's pure overhead, like there's no value to it. No pun intended.

A

Yeah, that makes sense, I mean the other question that happens like I think we briefly talked about this naming convention wise, but what I mean it was another one like video tuple like do we prefix or suffix, or do we cover something you yeah.

B

I, don't I don't have a strong opinion. I I, like the prefix, better I, understand the argument that making it a suffix allows it to show up better in you know intellisense and things like that, although if you type task and intellisense it's going to narrow it down to the set of things regardless of whether it's a prefix or suffix, so.

A

I mean it's mostly sorting, but I mean given that these things usually I mean these are types right I mean like for most cases. I would be more worried about because these are kind of tasks that you don't did so not api's. You use a lot yourself there they're created on your behalf right so you're, not instantiating tasks, necessarily that much so we'll be more concerned about how it affect signature, sorting and stuff like that, but that to me it seems more like, like itself all gonna like whether it's prefix or suffix, both work yeah.

B

I, don't I, don't have a strong opinion if you guys did I defer to you I, don't.

C

Have a soda prefer some things because to me that is about intelligence on it's also about buying documentation, and you know all there are many places where things show up border here, but I can't read it in this case, hi kind of don't you from I.

A

Mean arguably like for I, think for task I think we can all agree that this is not a an API. You should use a lot. It should be used very sparingly in cases where you know you want that. My hunches. That's probably true, for all of these cases, where we have struck based alternatives right like video topic for exam, what we think this is used by default or what we say well, it kind of depends well.

C

I, don't because of the cost, I very, very use them. So, yes, I will use the sofa.

A

W would say you would rather have it show up next to it, because you kind of want people to steer it to that. Well, that makes it I don't know right now. Should we want them to show up side-by-side or not? I don't know, I think we should just pick one pattern and stick to it rather than have two different ways doing.

B

I agree we should we have one convention, my slight preference and again it's not strong. My slight preference is for the prefix simply because it's how like I read the c-sharp, you know I read, see sharpens it's a struct foo or a class foo. This is a value foo versus a task value. I. Think the value is being like the T. The generic t result it's a task of int, it's a task of value, but that's a minor thing and it's just my mental model.

C

Another thing about feelings is we have many of these types and you you fight value. There will be very long list of those.

B

Let's say: that's gonna happen regardless, when you type out you, it's gonna you're, going to get a long list of things that contain the word value in them. So.

C

The thing is that you can find us you'll, not only peroxy cons, and then you can, you know, go up and down using oh yeah.

B

But I'm saying that's the same: that's the same either way but.

C

No because if you don't follow you, you may first time value, activator and value. You know, like intelligence will, show you many times. None of them are related tasks. They are just about Persians of some other us. It's not right has happened to be a you know, Francine automatically sorted order, and then you can scroll, but you have to keep writing to you. The letter T or you have to keep scrolling in the entire confusion with them. I.

B

Guess, I'm not quite understanding in either the suffix or the prefix case. If you type value you're gonna get a list of all the things that have value in them. A lot of things that a task in them.

C

Correct you have to know how to type.

B

Are you saying that you're just getting more likely to type the thing at the beginning of the word? No.

C

I think, basically, people are more likely to type the actual name of the time, not a substring of the time. So if they are looking for value task, they are more likely to type value than task.

C

B

C

So many cute on you, they will see lots of times. But if you guys like.

A

C

Want a more specific.

A

Thing first is natural yeah.

C

Like the thing that you know corresponds to what they are actually looking, if I over there bit, yeah.

B

C

B

Not seem to go against our like the standard, like stream versus streamreader.

B

If you want a reader for a stream.

C

I'm, just there are not that many times its size, jeans of the issue doesn't exist in my opinion, doesn't exist there. You type stream and UC fort I care, I. Think.

A

C

Just up and or down arrow to the one that you want, this is now you type value and potentially do not see one thousand times in five years. Yeah.

A

More like similar comparison, I think it's a physical out in the design guidelines are things like element node tree like things that are very generic right and videos. Hitler generic term stream is already somewhat specific source. Reader um I mean compared to the others, ah but I, don't know I, think I, don't have a strong preference, I mean, let's just leave it at that and then make a decision except maybe start an email thread and then closed it one way or the other. um So.

B

Summarize, where we are, we agree, we should add this not quite sure on the order of the terms in the name and not quite sure, on the assembly. Is that a summary thus far? Well.

A

Assembly, I think I think we are in agreement, it's just. We need to actually figure it out. I think the best I.

B

Guess, I'm not clear on how the assembly thing happens because, like if asp.net wanted to use this for v1, that would mean that it would have to be in desktop by then no.

A

Because, basically, we have a bit of a facade correct for desktop sure, and you can have like give it a separate type that facade can type forward everything to desktop, except for that one time, because that would then live in that facade assembly. So.

B

There's not it but right, but don't we have to doesn't that facade assembly get revved at the same rate as desktop.

C

Is all right, yeah.

A

Facade is just regular, dll right, so, let's get that gets deployed. I mean the only tricky part is. If that assembly that you're trying to add, already exists in desktop, then I'm not sure we can actually have an appt local. There was. There was a few issues with problem of how that works, but for the majority of cases that we have basically a contract behalf of it is forwarded to to desktop. We can a brand new types a base. You only live in that assembly we already attended for sim Nev.

A

You have done it in other cases. Okay,.

B

You guys know the stuff yeah.

A

B

A

For API is I mean if you'd like that, does this come of any other members, or is this basically just a self-contained type and that's it? It's.

B

A few salt content types so.

A

There's no additional members, you have to add to task, for example, no all right, then that should be relatively straightforward.

C

A

System threading asks.

A

Butter pecan bake this partial Assad.

A

All right, I'll.

A

All right, so let me so we'll just mark it as approved, because the point here is not to design the feature at the point. It's more like to approve the general fact that the API wants to be.

B

Edit and so what's the next step son so.

A

The next step would be we just basically figure out. A naming. A thing to me is back level work because it would just go ahead and leave it as it is for now, and then we can father back and rename it. If you feel we want to change to something else, it wouldn't block the work item on it. Okay, so I should.

B

So I should go ahead and check this in with either the names into the system. Threading tasks, assembly in core FX and.

A

B

Subsequently evaluate whether we can in fact update the partial facade for full frame work and with.

A

B

A

The other thing is uh Alex on West's team is looking into how to how to track work that we need to do for desktop. I. Think the idea is that we basically use labels on the issues, and then there is basically time review body that walks over them and make the decision was to do with those are the report them to the store, but we don't put them to desktop and if you poured it, how we port them, but I think this is something more like an architectural thing.

A

I would just ping West and say, like you know, can we make this a partial facade and if not, why not? And if we can't make it a partial facade than the question is, are we willing to come up with a new contract just so that they can edit to death, so I'll be living with effectively that it has to be desktop 4.8 or something because I agree with Chris stuff, like these Franklin designs, that we create these new contracts? Just so we can add it to desktop.

A

Is is a bit of a problem but I think.

C

You've learned, you know many features that we that, like logically, they are separate features. You should just be a contract. Yes, but you shouldn't, like you know, twist the design.

B

This is to be clear: this isn't I, don't have a strong preference. This isn't changing. The design of the types at all is just a question of where they live, correct, correct they sit on top of everything yeah. If we knew that we were going to add a bunch more things than we're. Gonna need a new assembly anyway, yes, but.

A

I think that I mean I, don't know I don't know, but if you have something in mind, that is that.

B

People just like I've had some we've had some requests for additional asynchronous synchronization primitives. We don't have any of them. You know concretely yet, but we had a bunch and blended in like the parallel extensions extras package, and there are some third-party libraries that provide such things. We've had some both internal and external requests for things like an asynchronous reader-writer lock. So you know if we were to going to go ad, something like that, we would. It can be done purely on top of existing surface area.

B

We wouldn't want to go put it in tennis court. Look, for example, oh.

A

No for sure no, but my point is that I think like a DS like, are they logically cohesive enough to say I, don't know system got threatened, got taxed, God, foo and then bae? Your task would go in there or are they all just additional things that yep some things may go in system setting tasks and some things may go into I, don't know threading whatever other things we have there and I think that's probably more likely right sure.

A

um So, anyway, all right! So let's leave it at that and these are poopy ruuu added. um Yes,.

B

So this one I I, don't know how to deal with PRS like this. So this one is it's not adding public surface area, but it is adding a publicly known interface to a type, an instance of which is returned out of a public API, the goal of which being to be able to cast a type that you get back to this other interface. So this particular guy has a method called search. His own method. That takes an I read only list and he wants to be able to pass a sorted list.

B

Skis collection or you know, sort of this keys into that keys returns an I list. He wants to be able to get it as an. I read only lists so he's proposed his PR. If you look at the files changed, it just adds the the interface implementation yeah. But to do that, you know, then, the code that you write for core effects will only work and certain you know only work against certain implementations and not against others.

B

So I don't know how to think through things like this.

A

Quite frankly, I think the value add is so low because you kind of have to know that you can cast it on. I would only this to begin with, and it's in a core type that changing is relatively hard because they have to touch it in at these three places, and on top of that, I'm not sure how doesn't have any civilization, in fact, I'm pretty sure probably does in some cases that I don't know I I don't I mean is.

C

A

C

A

Show keys apparently returns a list of T, so you.

C

Can change you can't change that, but he.

A

Wants to change the instance that we fit is internally in return. I, don't know what the incidence return. Even do, you know will be returned there if it's an internal type, but maybe possibilities public IP. It's a public API change anyway. Yeah click.

B

On files, change, you'll see it, that's the one like Isis.

A

Here it's a PR, yes, I see: oh it's a private time. Okay, yeah I, don't know I, don't know so.

C

You know what he could instead solve it by having do we have an adapter from a list or I read only list, because we could have an adapter instead of cast and then we would enable these scenarios and it actually would be safer because it would always say, Carrie cancel. You know, runtime accepted.

A

Ulok'neh right.

C

But yes yeah, but no yeah.

A

We didn't do that, it's for something you should have done. Oh.

A

B

God will have that for the concrete list, but not for I list right like then, we have lists as read only or something like that.

A

Yeah, but that I think returns a read-only collection, I, don't think if it rains in an interface. So we don't have problem with the only interfaces you couldn't make the existing one inherit from those which would be the same thing to do, and there's no adapter between the so you basically have to wherever you implement. I list you can't have to implement. I would only listen, no as well.

A

Unfortunately, that was one of the biggest problem that the design had.

B

Actually, it looks like I mean this doesn't solve this exact problem, but does look like we'd only collection implements. I read only lists.

A

Yes, oh I, see what you're saying now, yes, I think, there's probably some cases will be by you, but you could use an existing type as the adapter. That's probably true, but we also have I with only collections of the only thing you have is an I collection. You may not be able to easily wrapped it in another. One I.

C

Wondered whether you on the collection, I, don't remember.

A

C

Doesn't come no.

A

They are basically just views over whatever you pass. It.

B

C

B

It's called new read-only collection, I. Think.

C

They actually took a point if you much better than you know doing this cuz that will potentially failure on some platforms. Well,.

A

That's the other thing right, I mean first of all, yes to you have to knowledge, you can do it. Secondly, you would always have to be defensive, which basically heats the purpose yeah all right. um Another.

C

Thing is, maybe we should consider adding an expansion metal genre. Basically, you know, as you don't leave I'll slc. Oh.

C

Okay, so yeah, because it will be done, a location that it's not very visible.

A

The other one there were three collection pipes: artist I will only list I.

C

Collection I listed what there.

A

Was the all I dictionary yeah.

C

A

Say it won't always wake electioneering.

C

A

A

This sort of list.

A

Jesus so I'm doing this Rockies.

A

I'll try to block my typing. It's not usually that bad all right.

A

Anyway, this is me move anything you see on the screen.

A

Sorry, sir, so what I would you see on the screen? Any major disagreements? No, it's awesome all right, then it's mug disease.

A

It's not an API, but I think we ought never rejected part. Okay, then it just closed it.

C

Okay, that's probably the best way.

A

All right um so.

B

This one's just been sitting around for a while.

C

So, what's the main you should be, that sounds like a good idea at.

A

Sorting prisoners to be the rain builder.

C

A

That's another thing called it in the dcl, because we have two ways to do comparison so that you can either have a nikon para or you can have an eye. A comparison delegate. We editor in 45 the ability to create an icon. Para form a comparison. But again you I locate right. Oh no, new.

C

Colony the prison well.

A

That's the only problem for positive and I compare, but if you possibly delegate delegate the sketch for every single time right in the kids, I'm not sure how these to translate but I guess it's like I think my as far as I know the. If you do interface the stretch in the tight loop, they host it out. So it's only a one-time look up because.

C

A

C

Do the same conversion, it's.

A

Not like I was.

C

Sorted very complex, algorithm, yeah.

A

Like there's something I, don't know how they compare when you're actually pause in a comparison, resistant I, compare I, think for practical people inc purposes, because it's so hard to actually do the translation from different comparison to I. Compare you have to know how to do it because you go with compare of T DAW create, compare like it's, it's not great, plus, it's so common I would say they're having an overload generally. It's a good idea, see them with your opinion on that I mean I.

A

Don't think that we wouldn't add those seems pretty simple.

B

Yeah I mean I. This is a completely standalone assembly. We can always choose to do a more efficient implementation later, rather than you know doing. The rapping yeah.

C

B

Don't have any problem with it. Yeah.

A

I'll be happy with the implementation I mean it seems really straightforward, but yeah.

B

I mean my feedback to andrew has been you know. This is allocating and his response of well, if it's an issue we'll fix it later, which I think is reasonable,.

A

A

All because he basically does what I just said exactly I see yeah yeah, I, guess that makes sense. I mean, I think, imma just movies nudity standpoint. I too, like the API by.

C

The way did happen, not only other place it, it is an interface dispatch and then the implementation of the interface also close to that I guess: Ryan.

B

I sure it basically doubles and I'm prevent interface calls yeah. It's.

C

Pretty bad but fine we can organize.

A

A

Cool then I think we're good to go anything else you wanted. You wanted to look at because it's the only thing you sent me and in email those.

B

Were the main ones that I had top of mind? I, don't believe there were any other PRS that were blocked on you, but we might take a quick pass through there's a whole bunch of PRS like half the PRS are waiting for feedback from Richard on kind of the copyright header, but hopefully we'll get word on that suit. Yeah.

A

I thought we made the decision on that when I he said was mine um I'm, so.

A

A

I think there's only one: that's currently marked as API.

A

Additionally, some the pr side.

B

Yeah there I think there are a bunch on the issue side, but we don't have time in 15 minutes. We might look to see if anything jumps out at us, but yeah.

A

The issues that club is terrible- I came- I really really hope, I guy thing about cast a bunch of meetings over the next week, so I can actually sit in my office for a few hours, because I think most of it is just me reading stuff and responding I, don't think there's any any other way of doing it. If.

B

You could do that that'd be awesome. Yeah.

A

um Let me do a quick son here: labels API needs work, or maybe.

C

A

Now this 11 open right now, so that's not bad. Actually, I.

A

Guess the ones are more interesting are the ones that don't have that day. Everyone yeah.

B

We haven't been doing a very good job of keeping things attributed appropriately.

A

M9 to be Birds 27, but yeah.

C

A

All right I will take a pass I. Think honestly, most of these things like blossom, we went all that lists. The problem is they're all in the similar category, as in one of them are bad. None of them are amazing enough to justify you know all the ceremony on designing how we edit a desktop making decisions with them and I. Think most of our work is kind of block to have the same passes around that because I mean most of them.

A

I really just want to say yet looks good, looks good, looks good because they're all like adding value of somewhere. It's just that the cost is so high to actually add them. Then it's always a judgment. Call is that particular one nor off the cost of doing it, and it seems like a bad starting point, because then you're, basically always start with minus a thousand points which, for language features, may be worth it, but for simple API editions.

A

I think the cost is just disproportionate to what the value is of the API is all right, then I think I have it for the day great.

B

A

You much thanks.