.Net Foundation .Net Community Standups, 9 Jul 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Languages & Runtime: .NET Community Standup - July 9 2020 - ARM64, Single File, and .NET Standard

Description

Join members from the .NET teams for our community standup covering great community contributions for Framework, .NET Core, Languages, CLI, MSBuild, and more.

A

Hello friends of the net, we are a life of another version of the dotnet community, stand-up, focusing on language and runtime, and today we have a few interesting topics. We will talk about arm. We will talk about update on sonnet standard and then whatever carefully and Philip. Will you know interject at any given point in time? We'll also talk about. So how about you introduce yourself which start.

B

haha Stop stop typing.

B

Rich work on the runtime and other other parts of the platform and I'm excited to be here. It's been a while Kaplan you're next yeah.

C

And hello, how are you doing today and we definitely would like to get some questions from you about what you care about? So, if you're listening in today, I'm Kathleen Dollard I'm, a p.m. on the working with net SDK CLI languages templates can all over the map. So, like all of us here, we can answer questions about a number of things. So someone get some conversations going today.

A

And then we have two Phillips one of the needs introduced himself. Why.

D

Yes, hi I work on languages, compilers and tooling, for languages and.

A

Only this little Philip can talk right is that right, yeah.

D

That's how it works all.

A

C

D

It's pretty easy to change. You know, thanks to Microsoft, Corporation team's trademark, okay, its.

A

D

A

And then the Kaplan is no back again. So, like last time, my name is Alan. Roth I am a p.m. on the dawn, a platform team, so I focus on class. Libraries don't need stand under, we briefly talked about and whatever else comes up, usually so rich. Do you want to start talking about arm.

B

ah Yes, yes, I was just preparing. Let's, let's see if we can actually do this in a sensible way. It's preparing some links to share here so yeah, so we so this is about arm sixty four. We have kind of two projects in the hopper right now related to that and one that we've done so we added support for Linux arm 64 with Rio I. Think I should know these things that I worked on them, but and then with 500, even adding support for Windows arm 64.

B

So that's for, like you know the surface pro X and other similar sorts of devices like I, have an Asus device. That's like 50, centimeters away from me right now that I use for my testing here at home and yeah. Let's, let's talk about Windows arm 64 first, so basically our our plan for arm 64 was kind of twofold for windows, which was, let's make sure we can offer all the functionality that is available on x64 windows, and so that's what we've been doing.

B

Wpf isn't there yet windows forms is and I've been using it on my my Windows arm 64 box and pretty good, like I. Don't really actually notice much of a difference compared to my like x64 or laptop laptop class machines separately. We've been doing another project which is like make arm 64 much closer to parity with X 64 from a pure chip.

B

Standpoint and I'll find the link for that in just a sec, and that's one of those like asymptotic asymptotically, approaching parity types of topics, but we've actually made some major gains in FIFO, so for people that are actually running on arm 64, they should they should see some big wins separately.

B

Apple actually announced that they're moving to their own arm-based chip for their new laptops, which they're gonna first start shipping later this year, which is you.

C

B

Had kind of been predicted for a long time, apples been investing a ton in their a a style chips in the iPhone and the iPad I think they're up to a 12 now so the like. This is like the 12th generation of that ship. You know they bought a company like 10 years ago, or something like that and they've clearly made that a real thing.

B

So we you know various teams at Microsoft like office. You know, presumably, teams like vs code and the dotnet core team and xamarin. We all have to like figure. Oh we're going to do, and obviously we're gonna support the new Macs and.

B

Yeah so yeah, so we're starting a project we have dev kits coming to Microsoft they've, actually mostly arrived already and there's two kind of ways you can run code on these laptops. One is actually I. Don't one thing actually clarify, we actually don't know their laptops. Apple hasn't specified. What the form factor of the the first set of new Macs is gonna, be that we talked to them. They're, actually quite clear on that to us I'm just saying I think there are laptops, but I don't actually have that information.

B

That could be anything, and so anyway, you can run emulated in this thing called Rosetta, which will run x64 Intel code or you can run natively with arm 64. So anyway, that's that's. What we're working on our goal is always is to make it so that you can run dotnet code into places you want, and this this falls into that so I'll share some links with the team. Oh, who will then share them with with actually maybe maybe I'll start passing them to him. Let's see what would be the most.

A

So when you say, let's support the new chip like what-what, can you outline what kind of work you would have to do because I think it's abstract for most people like what did actually, yes,.

B

Okay, let me bring up my other team's window and we'll do this and okay. This is the right one, I think so. Here's the tweet now I'll find get ahead of you. Okay, yours did.

A

You get that yeah, so do you want to present them or how should I didn't.

B

Know you can do it, you can do it. Okay,.

A

B

Right yeah, so III try and be, as you know, public and transparent as possible about the work that we're doing so. I did a tweet that kind of showed, though the work we had to do a bunch of it is with these there's a kind of two major aspects: one is or there's a few aspects. One is as you'll see in this week there's this concept of heartened runtimes, which already exists in Mac, OS. Sorry, like x64 and now they're going to apply that to arm 64 all the times.

B

So we already had support for hardened runtimes, but now we have to make sure that support is kind of polished, that's presumably, okay, yeah. We have to make sure that work is polished. The next thing is: there's like this page model, and here I'm talking about like memory pages and how they're marked so in some cases, the done at runtime relies on, read and write, execute pages and those are not acceptable in this new environment. So we have to either have read, execute or read write, so this is called write exclusive, execute pages.

B

So we have to do work for that in this blog. That I think he mows bringing up now is from it's actually from five years ago of this same work going on in Firefox. That was just a good blog post that we found out about. So it's largely a security feature so and then the last thing is for some scenarios. Apple requires Universal binaries, which have both x64 and arm 64 code in, and we don't support that today. So we're we're looking into that. So that's kind of whole thing, yeah, I, think I'm.

A

I'm having trouble getting the thing up, but it will eventually come up.

A

B

Yeah ice-t that so, hopefully other people do.

B

ah Oh yeah on this particular thing right: exclusive, execute the windows team has actually asked us to do this in the past and we hadn't previously funded it so we'll be funding it now and presumably bringing it to all platforms. You haven't quite made all the decisions yet, but that's my expectation.

A

A

Okay, so I just learned another thing, apparently I mean something I share myself of teams, then I can't just share the team's window of OBS that doesn't work so again. Next time somebody else should share cuz, that's the same problem with the camera right, OBS or team doesn't show you or on camera. So anyway,.

B

A

Now have the blog post up yeah.

B

Yeah, it's it's Firefox in their JIT. It's not just a JIT problem. It's actually from us more of a runtime problem than a JIT problem, but same difference right.

B

I, don't have anything more to say about this particular blog post. It's just a good piece of context. If people want to read more about the general problem, space and I'll just share this with you, too is I. Don't know if you're going to be able to show that, but Apple has a document that we're using. That has much of the information about.

B

Their guidance- and we talked to them specifically about this document in China, yeah and and so I think Apple wants the the phrase Apple silicon to be used more than arm 64. So that's just when, when you see us blogging about it, we'll be using that term, primarily.

A

So is that the same as I'm 64 then like? Does it mean that our work actually works only for Apple, or does it work for anything that is arm 64.

B

Yes and no, the short version is, is pretty much. Everything is armed, 64 standard like standard armed, 64, stuff, the they have an API that other operating systems do not have, which is documented in this page. That is an OS level concept that we actually think is is good that we will be using. That is apple specific and we will not be able to use on other operating systems for our implementation of right, exclusive, execute, but other than that, like the other thing is, these are actually armed.

B

64 eat chips which I not not too sure those were used elsewhere, but will eventually be implementing some additional arm. 64 instructions as well, but not not in v1 anyway, I think we've kind of spent enough time on this make sense.

A

All right so um yeah I want to quickly talk about baby I've got a standard like a while ago, I published a blog post I think it was August last year where I direct some of the numbers from new, better work to see how far we have a for adoption and since I'm.

A

Now in the midst of writing a blog post for what happens, and one at five I was also basically pulling the latest numbers, and so I thought I share the numbers with you and also talk a little bit about how we collect these numbers, because I think the blog post will probably not go into much detail on this one, but it might be interesting to people. So let me just share the screen really quick.

A

So, basically, where we are right now is we have when we look at you get to work, we basically look at packages that are not owned by Microsoft and we also exclude packages that are not library. Packages, like you know, they're still, you know javascript-based NuGet packages that you would install into a spin, ed apps, for example, and so if we only look at the old stand in the top thousand seventy-seven percent of.

C

A

Dotnet standard, which is much higher than what we had in August I, think August was I, want to say sixty percent and then, if you look at all packages that are active, meaning them were updated within the last six months about sixty percent of them support on incident, which is quite a bit, and we look at all packages on you go to work, then it's only forty percent, but that's also kind of expected, because you know there's a long tail. Writing on all packages are actively maintained and it's not all packages would benefit from this.

A

The numbers by themselves are not always the most interesting one. Sometimes you didn't need to look at an actual graph, so we also have that we have a plotting over time. So, for example, you look at top thousand. You can kind of see that we have this. You know asymptotic curve now we're. Basically what it tells us is that we cannot really expect growth anymore in the top thousand, so we are about eight hundred packages in which kind of makes sense right, because not all packages, and even at the top thousand, are actively maintained.

A

So that's kind of what you would expect I guess so I think. As far as twelve thousand goes, you know close to eighty percent. I think is fine, and then we look at the longtail. This is basically all packages right. You see more of a power curve, kind of thing or exponential curve.

A

This is little lip here with its plateaued and my first inclination was to say like okay, we messed up our data collection, but talking to our data guys, it's actually not the case, because what we do is we ingest you get to the org data at the time it was published in New York. So, even if we mess up our collection, that should not impact this, and so the running hypothesis is.

A

If you look at the dates, that's kind of like the beginning of 2020, and so some people say maybe that has to do with Cove it that people had other issues or vacation or some other stuff that that was out of the ordinary. But it's an interesting blip that we don't really have a good explanation for right now, yeah. Basically, the idea here is, if you look at the curve here, you can expect more growth right, so, like yeah I, think New York has total 250,000 unique packages, so we are about what is it?

A

Ninety ninety thousand packages in weaved understand and support, so that looks like there's more growth still to be expected, which again also makes sense right, because even if you look at it, you know last six months, there's plenty of packages that are only updated when somebody really twists the package maintainers arm, and that might only happen like once.

A

Every year right I have certainly NuGet package that I published you know ten years ago, almost and I only update them and I have to write so that's kind of a table of numbers, so I think you know all in all, I think it looks pretty healthy, massive adoption, which basically means almost everything that is a numerator or can be referenced by a net standard. And, as you probably know, we have this competitor net standard binary.

A

We will fall back to the network binary, but that of course doesn't mean that the binary will work all right. So it's still very important for us to see that package maintainers are actually actively targeting at center and do you know the tests and all the other things to come of it to make sure that their stuff actually works across all the net versions.

A

Any questions about that, maybe even when the people on the call.

A

No questions everybody's happy that scanned it's hard to believe it's.

C

Pretty good number, it's hard to say bad things about these numbers. Isn't it it's.

A

True I guess that does fit as a fair point. I.

B

Have a question which is and I wasn't paying complete attention because I was actually responding to questions in the chat, so I saw on the slide. It said you know it's, you know, top 100 of thousand excludes Microsoft packages which, which makes sense. But what is the process looking like for Microsoft packages because clearly people care about it? Those right.

A

It's so like so the way our data collection works is we use the sting cousteau and azure, which is I, think it's called data, Lake I think publicly I forgot my.

B

A

It's called cousteau as well: okay,.

B

A

Basically, like we ingest effectively the data from nuclear door works and people publish, and then we can run clear is on over that, and so what this tells us is fits effectively every single time. Somebody updates a package. We basically know you know what contents are in the package and we have all the metadata right, and so what this number means is basically for this query: we just excluded where the author is set to Microsoft right and so we're also looking at the Microsoft one.

A

The problem that we have is in the top thousand I guess is that you know we ship a bunch of new get packages that are effectively part of the.net platform right.

A

So they are very popular whether she effect that they're referenced by everybody and given me the part of the platform, they always maintained right, and so that's why it's kind of like weird to look at the O's, because they would completely bias the picture towards or everything is, super well maintained and supports dotnet standard right, and so that's why we exclude Microsoft from those queries to focus on the third-party library ecosystem, which is a little bit more interesting and but like at the same time.

A

Internally, we do have as a first party engagement where we basically work with people inside of Microsoft. There are not part of the.net platform rights, for example the azure SDK is, you know you can think of it, maybe as part of the.net platform, but it's not, but it's another team that basically builds an SDK for various languages, including net, and so we of course want to work with them to make sure that their stuff is available for dotnet standard, and that is going as well.

A

I don't have the numbers on those today, but that would be interesting number to look at as well, and we do track that internally that there's a there's a chart somewhere that we didn't maintain.

C

So there's an interesting question in the in the chat that I think is worth a little bit of conversation, which is when's V s, going to make the jump into visual studio, we're going to make the jump to net core and I.

C

Think it's in it's an interesting question right now, particularly since we've got code spaces out there, and you know we are looking at major evolutions in visual studio, we're the.net core people who look at what visual studio is doing, not the visual studio, people who actually know what's going on so I have to caveat with with that. As a start, that I think we probably still have a perspective on that go.

D

Ahead, yeah I would say that there are. There are things in visual studio today that 100% cannot work and will not ever work on dotnet core, and so those would necessarily have to change to be able to make visual studio work on Donna core.

D

What it's not like, a question of. Could that work be done because you know given enough time. We have plenty of talented engineers who could certainly you know, make some definition of visual studio run on da mic or what's more challenging is like you know, okay! Well, if you chop off this piece and then this piece and then this piece to make it work like how many customers are you losing because of that, because you know the those you know, people depend on some things, working as they do today and- and they don't want to.

D

You- know- wake up to find that oh there's, a new version of Visual Studio and the thing that I use every day is now gone. Oops.

C

That's not just that's not just Microsoft creative Visual Studio pieces. There's a lot of physics support! That's out there. We have an ecosystem and not everyone is using pieces of that ecosystem, but the people that are it's pretty critical to them getting their job down. So we also have this ecosystem, which is also invested in the current old way of doing things, I think we're making in steps.

C

So, for example, you know, there's there's going to be changes coming up to the template engine to make the new project dialog work better behind the scenes and that's going to wind up affecting some templates they'll be backwards, compatible.

C

But at the moment, they're they're finding the the soft spots that we can get the most value for changes that well, you will either not see or see only as improvements and that's that's kind of a the first place to work that, from a visual studio as a whole is just it's an enormous beast, and so it is going to be hard to get it all. On two on to.net core platform and as philip said, it's gonna be especially hard.

C

We have to look at things that we're doing today that we're hacking through comb things that we would have to do a completely different way. Yeah.

A

So they just on that know, like we also work with plenty of other first part, is what vs is just a very obvious one, because it's you.

C

A

A big one that we all use, but the ability is that, like almost everybody who has a big application, whether it's exchange or sharepoint or you know, vs itself, given that they're all fairly old, but they all like ten fifteen twenty years old, they pretty much use every technology. We have internet framework right. So if you poured the whole thing wholesale or if that's your approach, then almost that requires to be right of large chunks of it right. So almost all teams, including vs, then try this the other way around.

A

They say: okay, what is the next things pretty building and how can we make sure that those investments carry forward to core right? So somebody in chat already mentioned that in envyus, for example, already have things out of prague and so some parts that one out of proper actually running in dotnet core out of prague already today. So that means you know, there's some IP c-- going on where we exchange messages with these processes, and these processes can run on core. They don't have to run on full framework right.

A

But if you look at you know little studio itself, I mean those abdomens. They use remoting right. They have in prop designer stuff running still in some cases and so.

C

A

Is in there right and like so like moving all of the stuff is hard, and so one one major investment we have done with donek or recently was moving the WinForms designer from in proc and visual studio to out of proc, so that we can actually run the design on dotnet core. Even though visual studio is rendering the control right and those are the kind of work that that needs to happen to to to make more and more parts of vs run on core, because we can't do it all.

A

So we have to do it piecemeal. Somehow,.

B

Other stuff is that a couple thoughts on that which is you folks, a kind of sadness I just want to make it super explicit, which is you so we're we? We basically have a policy that we're only gonna have one runtime in the vs process, and right now and now it's here- we're talking about dev MXC.

B

So right now that's done a framework, and so any dotnet core uses like, for example, the windows forms designer. It then necessarily needs to be out of proc, so this is kind of what emo was getting at, and so, if ever does go on top of Donna core, then the whole thing would switch, which is Donna core will begin proc. It would be the only in coop proc runtime in Devon, and all the data framework scenarios would then become out of proc. Then this kind of ties them with what Kathleen was saying so done it.

B

Let's so vs does use a lot of calm as like as its traditional model for componentry and done a core actually has fine support for for comm, so I can totally do that. What the challenge is is this that calm API assumes in process calls, and so, if some of those things that today come either call or are called as calm, api's Nevada process, then that's that's. The fundamental issue at play is that unless you build some kind of transparent transport which is probably impossible, then it's not the really.

B

The fact that they're calm, that is the problem, is that they moved. They moved outside the process, so yeah.

A

I mean that's usually the problem is that, like you know, chrome had this idea of transparent, like boundaries right and then very often doesn't necessarily work. When you make assumptions on the receiving end that you're in the same process as DUI control and so yeah, that's that's kind of the work but rich. You said something that is interesting. You said we made the decision that in vs you will only have one runtime. We generally don't support.

A

You will run time activation anyway right. We don't know.

C

A

C

B

Yeah good point and.

C

Yeah, you can run into these these boundary crossings, which are a little painful right now in odd places. So, actually right now we have a visual basic issue still open on the wind forms designer, because the way that you have to when you double-click on a control, you expect to go to the existing code for that control. Well, that involves several calls back and forth to find it in Visual Basic, because it can be in more places than it can be as easy part.

C

So it's a more somewhat more complicated solution there and it requires some hops across that that boundary and obviously that's not desirable to put a lot of those in. So we're still looking for a better answer for that solution, because sometimes it can be a challenging problem when you have an out of you know and out of park piece and where you're jumping back and forth across these boundaries into things that are like really Visual Studio is so welded to your source code, more so than I. Think probably anything else, that's out there. It's!

C

It is your source code and that's improv code is in proc in the in the the win32 framework, part of Visual Studio, and everything else does to talk through that and that can be challenging. So.

A

One question in chat is kind of like you know: how does donate core support call rich? Do you want to talk a little bit more about what that means and how it works or what is supporting? What isn't.

B

You look so I think from what I understand so I've been working with folks like Aaron, Robinson and Jeremy Kay I can't remember actually how to pronounce his last name and from what they've led me to believe. We support the Duquette core supports most of the things that dinah framework does so, for example, you know it supports in proc and out of proc managed calm servers. It actually has some features. I think that dotnet framework doesn't even have in terms of like the calm interfaces that it supports. It supports most of the same ones I.

B

My understanding is that dotnet cores I dispatch there's some something around this vicinity. If I dispatch, it doesn't do the same thing as Donna framework, so I think that is a feature gap, but for the most part it can do the same thing. So, for example, like I, don't know two years ago now there was a demo created of like a dotnet core console application. Automating office. I'm, sorry excel is what I meant to say. So most common scenarios work the I dispatch one and then I could be slightly using the wrong term.

B

I think I'm over.

A

The only thing that I am running into is it because I have a console app that automates, Excel and apparently the API is I call require parts of the dynamic language one time the DLR part for dynamic support. Apparently that doesn't work on core yet think.

B

I'm sorry I remember also seeing a conversation on that not not so much Excel in particular, but a dynamic thing to remember that being added in five, though you haven't tried.

A

Have you tried that on 500 I think the last I tried must be one I believe, okay,.

B

A

That that was added in five. Oh, that would be cool core yeah I mean I mean some people seem to be confused, that it works in principle, but that shouldn't be super confusing because don't net core has windows, only features right by heritage and yeah. If you try to activate the comm control on Linux, that won't work right, but that's the same way as passing a window style path. The path API is when you're running on the next one, if you say C, colon backslash blah.

A

You know then also won't work on Linux right, but that doesn't mean that we can't have comm support on Windows. The other question I had was so what happens with calm activation, and you know office, loads and editing right on, because when we write under net framework, they can activate it on a fabric, one time, because it's centrally installed in the box. How does that work with we've done a core based seven things. Oh.

B

Sorry I was only half listening, but I think I understand what you're asking you're saying if you, if you expose a manage comp component and it needs the runtime to to load, what are the rules around that? Well.

A

How does it work yeah? Basically, that doesn't work at all and, if so big, how does it work here? Yeah.

B

It does work, we yeah, we worked on this and it has a slightly different policies and loading applications so with applications. The rule is that we roll forward on servicing versions so actually backing out. The major problem at play here that we're trying to address again is one runtime in the process which we just discussed with vs and so yeah, so for applications.

B

We roll forward on servicing version that doesn't make sense in this case, because you could have multiple runtimes that were made for, like you know, one for 301 for three one, one for five, oh and I, so I think we made. We either made the choice that, by default, its latest major, much like dinah framework traditionally was or it was, or it was latest latest minor version for the major version you used or something I can't remember the exact rule, but it basically rolled forward farther than applications do except that's one aspect.

B

The other aspect is.

B

We only we only support shared framework, managed comm components, and this is for a very similar reason because, like a lot, some people would like self-contained ones, but then, by definition, it's it's gonna, be the self-contained runtime of like one of the comp components that you loaded into the process, and we actually had a fair bit of trouble. Reasoning about that.

B

So we decided not to support that scenario at all. So and not everyone likes this, but it assumes it will only load the runtime from a global location.

A

B

It kind of makes.

A

Sense because I mean once you do self-contained, you may do trimming or other things and then it's like well, we get like.

B

That or something.

A

Weird, like that's yeah,.

B

That's another actually think we got as far as discussing that point, but that's an excellent point.

A

I mean the kind of is yeah. It's the same problem. If you don't know that framework right, once you had multiple runtimes, the first comm component that gets activated basically gets to choose, gets to choose. Everybody else. Just gets whatever that person decided right, I, think I, remember like back. In the day we go to our own native shim that basically, so we had a policy.

A

The basis says use the latest framework, that's on the box, so that you know if some, if we get loaded earlier, that we don't regress, somebody else's come component, a targeted and later version, but only that only helped us when we were the ones loaded first, if somebody else could notice. First, they didn't have the custom shipment, so they were running on some other version and then our stuff would still break well office.

B

Actually had custom code and like MSO DLL, to like help the quote: unquote: kind of middle actor.

A

So it has been.

B

A

I mean think I think that might have been would be time frame when we did that so, like God knows so, that's nice. So anyway. Other question in chat regarding column just opens over that yeah. Some people sound like some of the stuff in vs, like language, service, compiler and mis builder, running out of proc already moving that to core would be improving perf and that's what I mentioned earlier, like some of the stuff already runs on core right. If you.

D

Not necessarily there's.

D

Something as massive as a language service, there's kind of no guarantee that something is going to be measurably better. Oh.

A

Yeah they're, just wanting to the so you're responding to the perf claim, not the claim. That's.

C

D

Like these language services are extremely complicated, doing a lot of work, and you know it may turn out that you know that the kinds of performance benefits that exist in core steel are just are just kind of like in the margin compared to you know, other things that you can do. Yeah.

A

Makes sense to me, like I I, mean that's the general problem of like people comparing perf right, so we talked a lot about the perf improvement, so I'm done at core and I. Think Steven is writing another blog post on that, but, like a good chunk of it has to do with server right, so it meaning it doesn't really apply. If you, you know, let's say app or your WPF app to core, that in and of itself may not result in a proofing provement, because it depends on where you're spending your perf right.

A

If we have not made WPF significantly faster on core. No, have you made winform significantly faster on core right, but if you do a lot of folly- oh, maybe that's faster, because we improve Palio but, like that's the same thing with language services right depending on what you do in a good chunk of. It is just compute that the compiler has to do with resolving symbols and that doesn't necessarily benefit from us, improving folly all right, and so that's a yeah. It always depends on the scenario.

D

That is I think most most most of us who work on this stuff do want to ultimately move to to Cork.

C

D

Like yeah I mean typically the problems that we have are not really compute it's well, it's really two things it's. The first is just allocating more than we need to that's nice, usually the biggest the biggest thing and causing the GC to do a whole bunch of work that it shouldn't need to do, and then the second is just doing more compute than necessary for certain things.

D

Like you know, there was this kind of long-standing thing that has been recently resolved, that we've been 16.7 and in rosslyn, where we were keeping a lot of old, syntax, trees, sort of resident and memory for some kind of niche scenarios that just so that we wouldn't have to recompute them. But it turned out that that that being stuck in my range was, you know, causing the GC to go wild for very large solutions, and some.

C

D

Some cases, because of Visual Studio to go home for little benefit to most users, and so it was one of those one of those things where, like you know that you know I'm. Switching the runtime in that sort of scenario wooden wouldn't actually help the problem at all. The the problem was really just sort of a design decision that that was like useful for some less common cases, but not as useful for much more common cases where you have larger solutions and like working out those sorts of problems.

D

That's usually where language service performance comes into play.

A

Yeah, but also that, because we have separate processes now, we have effectively independent GCS and we also effectively have for the awesome of the out of proc stuff. There's 32-bit memory limit doesn't apply because the order proc things can run in 64-bit and can use more memory, which also means you can burn more data, caches right and.

D

So, like you know there was this team threat from a Kevin pill, she's the lead for, like all things, asp net right now and I. Remember. He said that he had the entire ace peanut code base, open and Visual Studio code on his laptop, and he like left it on overnight or something like that and then, and and so with. What's important, there is the BS code model.

D

Is you know the language service runs as a set process, in this case a 64-bit process and when he opened up his laptop, it was using 17 gigabytes of RAM and the machine was just paging memory, the entire time non-stop, and it was basically like his laptop was useless, and so he was like. Oh well turns out, you know, 64-bit doesn't magically solve a problem.

D

It can you know, depending on the problem, it can just shove it under the rug, and you know right move it move it somewhere else, and so you know one of those things where, like you know, if your state, your code base needed like three gigabytes or something or 4 Giga or like just an extra couple, two gigabytes, then you know being 64-bit, would probably be okay, but like it's, not one of those sort of like blanket solutions that we have and.

C

D

Yeah and I actually be that very similarly to the move it to doc or thing like you know, in some cases, moving something to dot. My core will just result in a lot of things working better, but in other cases it won't really have any measurable impact and so or or it'll. Just like you know, you maybe have this other problem and it just kind of shifted it somewhere else and.

A

You might also make things worse right because the thing is like moving to kora holistically, very unlikely, making things worse. But if part of your stuff has to remain on full framework, then you would have to coordinate effectively now between the dotnet framework process in the.net core process and good example. Is the WinForms design aware every time you drag a control now or you interact with the control. You basically have to call across the process boundaries right in getting this thing.

A

Performant, it's not impossible, but it's of course work that we have to do to make the design I feel snappy right and so that that's also why, when people say perf, it highly depends- and somebody in shed rightfully pointed out to be compared perf all the time right and there's some truth to that. Right, like marketing has to happen, and we need to have supposed points on that list, but I mean we pretty much always say that you know purpose in the eye of the beholder.

A

You know, measure the things you care about and then compare the numbers for yourself because there's really it's really hard to say. You know things like cores three point. You know three times faster than full framework right, that's like almost all these claims are wrong, depending on what the benchmark is you're. Looking at.

D

Yeah and there's also on the linguist service front, specifically an interesting thing about moving out of proc like the more things you move out of Crockett sort of this point that you made about how come to coordinate data back and forth between different processes that can have an impact on someone like that there actually is a benefit to being in process or things like Diagnostics we're just getting very, very quick feedback on a change that you made in an editor, because you know the engine that sort of has to compute.

D

Whatever that change meant, and then you know, report stuff back to the user doesn't have to cross a process boundary right. It's already there. It's just ready to go, and so both you're now hopping between processes say. Maybe you have LSB and you're having that data being serialized and a JSON and then do serialize back into you know some objects and even stuff like that.

D

Like that it, that is a cost, and you know it's not necessarily I would say on the whole, that cost is worth it like I'm, personally, of the belief that all language services should be on process, but it's also it's nonzero, like you know, moving out a program will also have you know those sort of like little usability things that you you may take for granted today, where you know, there's kind like some.

D

So many of your changes, you're, like oh, it kind of, looks and feels a little different to me and that it's it's unclear if that's like positive or negative, because I think it just depends on the person and what they expect out of their tooling, but yeah. Those things have to be considered.

A

Big enough consideration and complicated choices which wanted to talk about Ubuntu 19 point 10 reaches end-of-life and July 2020.

A

You want to talk about that yeah.

B

So the thing I wanted to talk about, and actually now that I know that you have it up. I realize it was even another issue to share, but so we've been doing what I call distro management for quite some time and what that means is is like you know, we support like half a dozen or some number of Linux distros and yeah I just found the other one, and we haven't actually been like super transparent about exactly what that process looks like, and it's not like, not transparent in really a bad sense.

B

I guess it's like that. You know that thing is like. Don't think what is malicious for something?

B

That's incompetence, so this is definitely in the in confidence arena, and so we just haven't, you know, been telling people like you know this is this is what we're doing so I've been to 1910 is a version of Ubuntu that we we support and we're being like super public, saying it's going end-of-life on the 17th of July and then we have like all this like teardown of support or are kind of like you know, machinery and so starting 17th of July. That's what we're gonna start tearing it down until 17th of July.

B

We will support it and 17-2 lies their date. So actually I asked the team to start doing this kind of thing. The first one we did it with was Alpine. 312 assured that link with you. You know if you want to show it and so we're planning on having these posting these types of issues or both new new districts that we that we support and aol's like end-of-life ones as well, and our goal with the new versions is like. We don't have a promise of same-day support, but we're gonna, try and aim for that.

B

So we will like start so like a bunt to 2010. It's probably the next release that we will add support for so I'm, hoping, like you know, maybe a month and a half or two months ahead of time that there's an issue that gets posted Bourbon to 2010. To say this is this is what our process looks like and once all the check boxes are checked and that basically means it's supported. So I see that as a I move forward and kind of sophistication for the dotnet project.

A

So when you say issues like usually pulses in announcements right, no because they all in the quarry pool.

B

B

What we should do is there's actually a digest for product changes that is in the core repo I'll. Ask the the team to link to those.

B

Yeah I think these are a little too much noise for the announcements, although repo but I look for feedback on on on that, certainly from a checkbox checking perspective. I think it's overkill to have these in the announcements repo and like you.

C

Know we want yeah.

B

So I'd look for feedback.

A

On where people want these to be hosted yeah, this seems more like a work item, tracking thing or addresses announcements would be just okay, we just moved it out or we just edited or whatever right, yeah all right. So that makes sense.

A

So one question came up with machetes is dawn at 5:00, so that if I was only a few months away, is the feature set pretty much lockdown at this point or you're still looking for stuff to add, I. Think it's pretty much locked down at this point like we course still have some issues on the list of you want to finish. There's some rough. We haven't started yet, but we are not actively looking for net new features.

A

At this point, I mean, of course, exceptions exist everywhere right, but that's we want to ship at some point well,.

B

Like the only sort of exception to the rule, be something like Apple silicone.

A

Right like something that is, you know unexpected or whatever, but yeah I mean like we're still adding ap is right, like we're still reviewing community PRS and other stuff, but like we're not like hunting for new things, to add to the backlog.

D

D

At least on the language side, certainly for I, think for sea shark or pretty much sea shark man we're pretty much locking down now I mean you know, there's like little design quirks here and therefore you know the existing features that have shift and preview that that are certainly subject to no addition or change or whatever. But no, like you know, big news, these sharp features and then on f-sharp. We are looking to there's two more features, we're looking to finish and then I think there's another that we may add strain circulation it.

D

But that's gonna, be like a quality call like. If you know we have confidence that we could merge the feature and it would work for most people. But there's a question of.

D

If it's going to be, you know enough or or like good enough, I should say, or you know it says that we put it in preview or we just sort of like hold off and don't even merge it in at you know, behind a preview flag, we're kind of gonna, you know see, I, don't know just kind of see see what happens based off of how the current work is going and ultimately make your quality called.

C

And from an STK clr point of view, and in this field we still have things on the agenda, so we're still working, we started work to do and preview six does not feature complete. However, we think we pretty much have our backlog set. I will say that something was discussed over the weekend on Twitter that may make it in, but it's a very, very small thing and so we're pretty much set in terms of the way we're thinking about the release, at least from our perspective, we're not finished working on the release by any means.

C

B

A good time to tell folks what our closed down schedule looks like so yeah I believe on. In the first half of August I heard that I heard the date August 8th but I'm. Definitely not gonna promise that the first half of August working a branch for the 500 release, which means we'll create a branch for you, know 500, RC, blah and then master will become done at 6:00.

B

So that should happen in the first half of August, and that then that's a really, and that means also that our life process for getting features and bug fixes like that I shouldn't say features. There should be no more new features at points bug fixes. They all have to be approved at that point and go through a rigorous process.

A

So one question that was asked in Chad is what about mac OS on arm.

B

Yeah that'll come in servicing.

A

Servicing for five, oh yeah.

B

That's our current plan, like it hasn't, been locked down. Yet we actually have a meeting later today, so hopefully lockdown our plan there, my expectation so.

A

It means there will be no three one updatable support arm sixty-four. That's.

B

All so yeah good question. That's also. My plan is 500 only 500, plus only.

A

Yeah, so six would then be the first LTS that will support yeah yeah.

B

That makes sense, that's also what we're doing for Windows arm sixty-four. We decided not to back port that the 3-1, even though he absolutely had a conversation on that I.

A

Think we generally don't pick board features right that you only do it if there's a strong reason to do so. Yeah.

B

The windows are 64, one would have been definitely much easier for us to back port to 31, so that was purely a policy decisions slightly less of a cost. Although some costs cuz we're kinda a constrained where they kind of a hard period because of the pandemic but earth of Apple silicon one, it's definitely much more risk for a back.

A

Board and when you say servicing release I mean there's not a minor version of five right, so if I could will be a pet star update to five right, yeah it'll just be you know: five dot, o dot and probably more like a six digit number yeah, so that one question was asked is getting more on everything on the same vcl that was pushed to six right. That is generally true. We have pushed out some of the sermon conversion stuff that we want to do for five to six yeah.

B

Don't the only part where that's not true is we are doing the convergence thing for Azzam yep. So, as I'm uses the model run time, we are moving to the Oh yeah. We are moving to the new bcl, a.

A

New or right awesome yeah, it's yeah wasn't right now run someone over I didn't did for five. The plan is to get that on top of the core base.

B

C

B

C

That underscores that we're working on that the whole mono convergence we're working on all of that we're doing it on a couple of different in a couple of different areas. Right now it just isn't going to be quality ready for November, so it all got pushed out by a year we're basically taking longer to get it done so that we can get other things done in the meantime and so that we can, we can spend the time to get it right. So right.

B

C

B

Don't leave people with the idea that the Oise done part of the convergence isn't going to be quality in the photo conferences.

C

For shipping that.

C

So that we can keep the quality high on what we actually have out there, that would be a much better way to put it thanks: rich, yes, yeah yeah.

A

Yeah turns out, you only have two dials right amount of stuff, you deliver and the quality, and so oh, no one more one more dial yeah, but that doll doesn't work. We know that.

A

Oh, let me also be fixed time rights at this point. It's I don't know, buy like a no I'm, actually quite quite happy with what we have four five zero and we will not be bored for six or so that's also good.

A

Yeah, so somebody says it's not related to any of you, probably, but as unity game that I dream of the day I can use proper core API is in unity, so I've recently talked to unity. They are still on dr. insanity.

A

Oh, we try to talk to them about, and you know at least support in understand a 2.1, but that's still a few years out because they are on a fairly different schedule from from where we are so that will still take a little bit because right now they are still running on the Monell bits with their own runtime. On top of it, and so moving to some sort of converged, core will probably require us to do more of the heavy lifting on our side.

A

First, where we get, you know more of the mana workloads on top of core to actually make sure that they have enough surface area to plumb their stuff. On top of that, but I would think eventually that will happen right. It's just a matter of time, because I mean like what unity also wants to do. Things have nothing to do with just churning the underlying platform right so yeah later.

A

Kathlyn, did you have something to show or talk about.

C

I know I was just looking: I tried to install something to see if it made it into preview. Six and it didn't, and so I mean I can say that we're doing some work to clean up some of the dotnet new experiences. So we do know that they've gotten pretty out of sync with several things. It's think of it just as an eater.

C

You need experience when you say: dotnet helped, you're no longer gonna get the help you asked for plus the listing you didn't ask for and have to scroll back up to find the help that you actually asked for. So we're going to be doing some stuff like that to clean it up and we doesn't get a performance improvement, the dependency and the small things that are coming in to five and then four six and possibly extending to seven.

C

We have bigger ideas in that space and yeah feedback always welcome on Sakhalin, so I in the spill. Other questions: oh yeah, if you're a template, author I would love for you to reach out, because templates authoring is going to change a little bit over the next couple of visual studio versions. So this is largely vision driven by Visual Studio changes and we're trying to make uptake of that as simple as possible.

C

But if you have and we're working on backwards compatibility story as well, but if you have templates that you're delivering either to CLI or Visual Studio I'll, be quite happy to have a chat with you about what you're doing and how we can keep from making your life very painful, because you're supporting other people that we also care about.

A

Somebody just said: Scott Hanselman mood, colors I have no idea what this is in reference to, but.

A

About the dogs max is posting on reddit, the monthly updates. That's where I've received, apparently.

C

A

Right so then, unless there's more questions, I think that's it for today and as one of you has something more to add, we.

C

Ended on time, it.

A

Seems like that I think.

B

I got all my things in.

C

I hope to have more next month.

C

And taupe is muted, although we can see it's no I'm.

A

Good, alright I, both of you good, is the big one. Also good or does the big one want to speak for now or oh.

A

All right, then, thanks everybody for listening in and then I hope, I see you in I think what is it four weeks from now? Something like that right.

B

A

All right, everybody wave bye-bye.