.Net Foundation .Net Community Standups, 4 Jun 2019

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: ASP.NET Community Standup - June 4th, 2019 - gRPC!

Description

Join members from the ASP.NET teams for our community standup covering great community contributions for ASP.NET, ASP.NET Core, and more.

Community links for this week: https://www.theurlist.com/aspnet-standup-2019-06-04/
gRPC repo for issues and feature suggestions: https://github.com/grpc/grpc-dotnet

Suggest an idea for a standup: http://aka.ms/dotnet-standup-suggestion

A

That mean we're live now. Yes,.

B

That means what yeah.

A

B

A cast of thousands.

B

Okay, yeah so yeah. Today we have the G RPC team on which is amazing. I want to give you plenty of time. I'll. Let you introduce yourself when you start talking. Let me jump in with the community links at the beginning and then we can. We can switch over and- and you can do your thing so, oh, let me find there. It is.

B

Well there, it is okay, I said I'll, just I'll run through these we've got a lot of fun links today. First of all, Rick's drawl did a live reload middleware component. This one was really fun because I got to watch it kind of happen on on Twitter and stuff. He had some some great feedback. So the idea here is: it's middleware that, as you edit your site, the middleware is running.

B

Using.Net watch run, and you know he shows here updating things in the site, updating, CSS, etc, and you know it's updating on site, so he worked through. First of all, you know creating standard, middleware and just refreshing the entire like using a file watcher and just injecting JavaScript and reloading the HTML.

B

What he he found here was David Fowler, got back to him and said you're allocating like crazy, and so this was really cool to kind of watch his he he updated and looked at ways where he actually wrapped the response stream used. Some of the advanced. You know, kind of memory, buffers and span and you know, looked at looked at how to make it reload faster. So this kind of goes all the way through and shows his continent completed. Implementation of it also, then, towards the ends. There's some caveats here.

B

There's questions about you know making it work with http/2. There was also a question about blazer. You know: will this work with blazer and and how to make that work there? So a lot of great stuff here number one. This looks like really useful, middleware and then number two. You know we all learn a lot. Looking at this stuff, he found out with David's feedback on an allocation.

B

Well, this is a recent on net show. This was an interview. Shane did with team here, so so I'm, Glen and Ryan, and talking through some great stuff, so mention of G RPC, which we're going to learn some more about today. Also talking about the work Oh were you on there too?

B

Okay, the worker template and then kind of wrapping up towards the end with how do you choose between G, RPC and Web, API, so good, stuff, sure, heads continuing and I believe this is number 22 of 26 for for his series he's doing by the way. He's also going to be packaging these up as a book which is really cool, so so this has been fun to see all kinds of stuff in this v4 validation. So talking about validation, attributes, server and client.

B

One neat thing that he did to here: I, like is he talks about remote validation, so this is looking at client server for things like if you want to go through and say order tickets and you need to validate from the client before you know, showing it actually are there. You know tickets or products or whatever in stock, so showing showing how to do that. So lot of great stuff and I'm excited to see that book cool, so Dodi G here here. This is the updates, so I'm just pointing this out.

B

We've mentioned this one before, but here he's updated up through 250 samples now- and this includes 3, o preview 5. So this is a great resource for all kinds of stuff and again the he breaks these down very kind of bite-sized implementations. So most of these, you look and they'll they'll, be just very targeted and they'll show just one feature at a time which is really useful.

B

The e here going into asp net by a boilerplate, so we've talked about asp net boilerplate in the past. Here li has a over an hour view are over an hour video showing off.

B

You know just different features here, and so he just kind of steps through and creates things here, he's using an angular site and he's just shows kind of how to set that up so good stuff, ken with what's new and as your signal our service, some neat stuff here, including endpoint routing, which is cool, to see that, and so you know, dig digging through and showing the example of how now you can you can set up using the using endpoint routing and how that simplifies the code, so GFP C also.

C

Uses in point routing cool, it's it's a great way for us to provide routing and any anyone can write a framework and quite easily plug into it. Cool.

B

Thanks James, so so that has been really cool to see over time and the first implementation I believe with with to to was just turning it on for MVC and so now, endpoint routing lighting up a signal, rng RPC and it just kind of available everywhere for adding is really nice.

B

Well, I'm, awesome, okay, so here speaking of gr PC, this is this is from Chris and he's writing about using GRP services with server-side blazer. So I like here, the idea of looking at two new features and kind of tying them together. So here's you know explaining G RPC, getting it set up, creating a G RPC service and then he also creates a client which is a server-side, blazer client and then calls over to it.

B

So sometimes I'll find this is a place where, when I try and hook, two new features up the two features work in isolation, but the two of them too gathers sometimes you run into problems, looks like here. This actually worked pretty smoothly for him, so that's cool, so.

A

Just speak into that about like hooking up multiple new features, if folks haven't seen and I think you might have mentioned before, Dan's dock and build was fantastic, it hooked together, blazer the new service workers, your PC, you want to see all our new stuff work together. I'd highly recommend go, look at Dan, stop from build cool.

B

Yeah, that's you know a dance talk, everyone love dance talk and it was a really really really smooth talk. That was three really is there, where he kind of integrated a lot of different stuff and showed everything working together, smoothly and and I feel, like that also kind of showed the value of like here's, how all these pieces fit together to build a really nice application, yep.

B

Okay, this is cool a code stencil, so I actually contacted me separately about this and when we're at right, when we wrap up the community links, he's actually giving the way two licenses for free to people that are watching the show. So I I will put the link in the chat right as soon as I wrap up the community links and I'm, not sure if it's the first two responses or if it's a random drawing, but you can enter there for for free license.

B

So what's neat with this, this is a code generator it's going through and it will look at your database and generate code based on it. Sometimes code generators are pretty verbose from looking through the actual tutorials. Today I was I was pretty happy to see kind of the balance of it that it does generate. You know, model classes and those sorts of things based on your database tables, but and and also your views. But then there were other things, for instance like the navigation where it was pulling those it was creating that dynamically.

B

So so pretty cool and the this specific post is talking about the razor stencil. So this is razor pages generated from your database so and again, watch for the end of the right, wrap up these links, I'll be putting that in the YouTube and twitch chats for people to shoot, then, on that free, giveaway, alright bill writing about augmenting the I principle when using identity, server, authentication middleware in asp net core, so he's got he's using he's using the authentication middleware.

B

He wants to augment the the principal to add additional information to it and he found out he ran into a problem with which was that things were overwritten at the end of the like. After the authentication. I like this animated gif here that he's got by the way, so at the completion of the the authentication, then it was being his values were over it and see that here the property was null.

B

So this was one of those we're actually going into the comments. There were some some useful information here, so you know dominic, who helps run. There is one of the two authors on identity server said: well, you could, you know, create a PR, etc and khalid who's on the same team. A trim, dev said you know they're on, I didn't serve you at this point. The another nice thing here is Thomas is saying you could use and I claims transformation, and so bill actually goes through and shows here's some sample code to do that.

B

There's some small things that you lose as far as control there, but overall this seems like a for most cases. This would probably handle things here all right. This is a cool just kind of quick one here. If you want to start using ASP net core 3 preview, in-app service reminder that you can go through. First of all, you'll hit an error because it's not deployed on app service yet, but you can do a framework dependent deployment.

B

So just as part of publish you can set deployment mode to self-contained, set your own time and you're off to the races.

A

No one got to because see a lot of people run into when do self-contained make sure you pick the right bit. Nurse gives app service defaults to 32-bit apps. Unless you go into the portal and change that that's been.

B

A

In that screenshot, that's on your screen right now,.

B

It's here, but.

A

They need to match.

B

A

So that's a common pitfall. We've seen people run into and if you scroll past it, the tooling and Visual Studio also prompts you like, when you try to publish that the runtime doesn't exist in app service. Just yet so that's yeah we're trying to catch. You make sure you don't fall through the cracks then end up in a failure case. Good.

B

Good, so the point right there is that before even going out to deploy it's gonna say: hey, wait a minute you're trying to deploy in a runtime. That's not yet, and.

A

If you go a little further up, you have this so we've had you know the generic 500 point. Oh ever page news is rendered by the SP net core module, there's some fun stuff coming there. I can probably talk to it in like a future episode, but we're doing a lot more to handle startup error failures. So we log a lot more to the event log. When possible, we show details in the browser. So all the stuff is coming in three, oh and you should look forward to that. All right cool all.

B

Right all right, I'll, try and hurry through cuz I want to see all your great stuff. Here's a quick thing: here's just a tweet. This will be shared in our show links. But damien mentioned that the team is doing some basic compat testing for nougat packages with asp net core 3. So this is, if you want to vote, if you want to say, hey here's something, that's important to me. I would like you to include in testing when to make sure people saw that all right, this is fun.

B

Mahuri has wanted to look at taking a blazer, app and hooking up with electron net, so electron net to create a desktop application, so electron dot net has support for a global tool. So there's, actually you do dotnet to install electron net CLI an it on the application, make some small changes and then to run it. You just did electron I start and here a little hard to see, but this is actually a full application here.

B

So just kind of neat I know there's already been that prototype I think blacktron was the name, and that was one that the team had done early on. This is using electron, electron eyes, dotnet, which is which is neat or electron. Doesn't it.

D

B

Scott Allen here is talking about. How do you, if you need to know at runtime, if some, if an authorization policy will succeed so say, for instance, if you want to update your menus on a page and only show like say admin links to somebody that would be authenticated as admin and and especially if an authorization policy is running some logic. You know here, you stick got just a very simple random number one, but let's say you needed to do something more more in-depth.

B

So fortunately you can just inject the authorization service and then you can say authorization authorized user and that'll. Let you know whether or not in this case whether the user is lucky but, more importantly, whether or not the authorization would succeed. So this is nice to know whether you can you know you can modify your page layout. You can modify what you're displaying to user based on whether they could pass an authorization policy. And finally, wrapping up here we've been sharing these weekly links using different.

B

What was it one tab and we've used some other things in the past, so Burke, Holland and sessile Philip both saw that you know. We've got some kind of halfway solutions, they actually created URL list and they did it. They said they had our show. The dotnet me stand up in mind when they were doing this, so this is a neat walkthrough that they do where they're using they're doing things serverless as much as possible. So they're things are done using Azure functions or writing stuff out blob storage, cosmos dB.

B

They have a view and typescript front end, and the end result is that you can go through and create a collection. So you can do this now you go to the URL list, calm and you can create. You can create something so in depth here they walk through everything how they built it. It's out on github. If you would like to get involved- and here are our community links for the week, so it worked out just great for me- I had some feature requests because of course he'd do that.

B

But um here's all our community links for the week I'll be sharing these in the show chat and you can go through and also I recommend that you play with the URL list and with that I'm gonna, stop talking and I'm gonna pass it back over to you, folks, okay, alright, so we got the G RPC team. Do you want to introduce.

E

Yourselves, alright I guess I'll go first, my name is John. I am a Devon. They asked me my core team and I've been working along. My friends here on the G RPC support from SP net core, and we've been doing that for the past several months. So now we're going to showcase what we have hey.

A

Folks, my name is so noob sure hearty I'm, a program manager on the SP night team and I've been working with John and rest of the folks here and.

E

A

C

I switch to the team's fears that people do see faces, see James Kenda I, my name's James Dean and Ken I'm, also a dev on a speed on a team and yeah John and I have been working together with the GOP C team. At the cloud native computing foundation to bring GOP C Donna core and from the CN CF team is yarn who's on the call with us.

D

Hey so I'm young I'm, a member of the GRDC team at Google, actually and I've, been on the GRDC team since JRPGs early beginnings, which is I. Think, like four years ago now and I'm, the owner of the original ERP, see sharp implementation.

C

Cool, so to start with I'm gonna share my screen.

C

Can someone handle switching to it? Yes, I.

E

C

Your slides, okay, great so to start with we'll just do a quick, a quick little slideshow to sort of introduce GI BC to a lot of donair developers. It's something which I don't think, there's been a huge amount of mindshare around and done it space, but GBC. It's a remoting, well RPC framework. That's caught on quite well when, as far as we can tell as the largest mindshare out of RPC frameworks like there are some alternatives out there, such as Apache thrift, GOP CX comes from the cloud native computing foundation and geodesy.

C

Its is actually a recursive acronym. So if anyone asks you what it stands for, it actually stands for. G RPC remote procedure calls G. Rpc is built worth modern technologies. So it's it's a relatively modern RPC framework. It was sort of open sourced around 2015 and it uses HTTP 2 for the transport and users. Various HTTP 2 features such as streaming, and it also uses protocol buffers in a number of different ways. Although proto buffers usually quite closely a soda associated with your PC, you can use other other formats for sending messages through it.

C

But we'll just be focusing on protobuf today for simplicity and GBC. It's designed for modern apps that has quite good support throughout all libraries for built-in tracing and logging, and also technologies, like deadlines and cancellations, to ensure that your overall application infrastructure doesn't have a runaway process. It consumes lots of resources.

C

So protocol buffers it actually serves three purposes with GRDC, so the first is an interface definition language.

C

So we can see here an example of a protobuf file and we're defining a couple of messages and then a service called topics, and then it has a subscribed method on it, which takes a request, and this particular method is a streaming one which will stream from the server to a client and it will stream multiple event. Objects.

C

um Protobuf also serves as a service model, so I'm generally cogeneration, we'll run over the top of this protobuf file and then generate your service CSV file, a client for your T RPC service and also messages and then finally, protobuf acts as a wire format. So when you send messages between the client and the server, your message will be encoded and protobuf, which is a very efficient binary format.

C

E

Actually, James I know I've seen some questions from the community about protocol buffers, and one of them is actually when you use jar PC and you want to use something else other than protocol buffers. How would you go about doing that.

C

Today and well, there's there's many different frameworks for Jia BC, so obviously it will depend on what framework who, using with the one that we are developing. There are some ways to hook into the serialization methods.

C

It is possible I, don't think what have time to go into it in detail now, but it is something we. We would probably like to support people to be able to extend done then and considering their own serialization formats, but it probably won't be like our focus. Our focus will be on delivering a good protobuf experience first and then more extensibility in the future.

A

There's a community project piece: you don't know that uses message back, but there are. There are people who living life right on the edge who are trying things out. It's we won't get to it for the foreseeable future. Yeah and.

E

A lot of the works that we're doing in terms of like optimization will also be more geared towards protocol buffers, and that will be our first class experience. And then we will potentially looking to other accessibility later mm-hmm.

C

So a key benefit of GI PC, this HTTP one and wrist and Jason is its performance. So one way it achieves that is through a low network usage. So it should be it's a binary protocol compared to HTTP 1, it supports header compression, so header compression becomes really useful when you're sending commonly occurring headers. So, for example, if every G RPC request has the same content type, we can use header compression to convert that from some utf-8 text down to just a couple of bits compared to HTTP one and also protobuf.

C

It's a binary serialization format and it's it's very efficient compared to JSON. So if we compare a JSON message versus a protobuf definition, so this is an apples-to-apples comparison, because if I was actually showing the protobuf message, it would be just some some random characters, or maybe some basics before uncoated text. But we can see here on the on the left Jason message: some things notice about it, as it includes the property name for every single property, so there's size there and also its self describing so we need to double quote things indicate their strings.

C

We need to have open and closed brackets. We need to have semicolons to delimit everything protobuf, it's a it's a binary format which relies on you having the definition to be able to make sense of it, which has some cons, but one of the very big pros is it's extremely efficient? So if we compare the two, we can see that protobuf is probably a third the size to a quarter the size of Jason.

C

The difference would shrink. If you were using gzip, it might only be maybe half the size, but a protobuf message will always be smaller than an equivalent Jason message.

D

Yeah one thing: I: would it's also much more efficient to virus and to Rose ice yeah.

C

That's a good point: the pets that the binary format that it has lends itself quite well to submission serialization compared to comparing characters and Jason. So.

A

We've got a question on Twitter saying: can you use this in that Nick or do two or is it dependent on features in three? Oh and.

C

A

C

We are making will be dependent on three. Oh, there is a a different gob C framework, the existing one which you might see on the tree RPC website I, believe that will work with two I.

A

Think one thing we can just you know clarify terminologies that we can disambiguate when we use the term jar PC core, which we will use multiple times later today, we're talking about the existing implementation, just based on your PC's existing C code implementation, the new stuff that we're doing is based on our own networking stack, so GRP C, the C core based version will work with two to the version. That's based on gastro needs to be client will not work with due to.

C

Perfectly answers so another benefit of eight should be to is multiplexing, so we can send multiple calls via a single TCP connection.

C

So this is a comparison I recorded on my computer of a image which was split up into two hundred small images, and you can quite easily see the quote: enormous performance improvement by being able to use multiplexing.

C

Another key feature of ERP see is cogeneration, so cogeneration is at the core of G RPC and it's driven by the protobuf IDL.

C

So, given our protobuf file that we had before we had a messages, I mean our service with our method. What we would do is include that prototype file and the CSS proj. So it's if you can see my cursor it's this protobuf line here. We just include the file, we say what we want to generate. In this case.

C

We want to generate a server stub, but you can also say you want to generate the client or maybe both and then the compiler will automatically invoke the koechner and generate a service, stub and messages from this protobuf file. So this is a very, very simplified view of what you would see for the service stub. So we've got our method, which takes a single request. It streams a response of event, and then it has a server call context which is kind of similar to an HTTP context and dotnet, and then you would take this class.

C

You would override this method and then you would hook it up in your setup file.

C

It also would generate a client, so this is a strongly typed client which would handle serialization and building or a correct URL, for you.

C

Another key feature is streaming.

A

See you got another question coming in on twitch, which is does do this? Is their mobile support for HTTP and can they use it for mobile devices?.

C

Yes, you can so I believe I believe what we're writing should using using HTTP client. So what we're going to talk about the client? Obviously, because it wouldn't make a huge amount of sense to host the server on on a mobile device, but we're we're creating a client which uses a tree. Rpc client, which uses HTTP, client internally and I, believe that should work on mobile devices and xamarin I know the C core client does. Is that right? Yeah? Yes,.

D

Well, it's it's been used by some users. We still call it like experimental support, because we don't claim it's like fully supported, but I I've heard from users that been using that successfully owns a marine okay.

A

And to that end, if you don't want to use the Ameren, you can still generate clients for I, think objective-c and Android both supported so you're free to use whatever ecosystem. You already use on your mobile devices and still talk to a Joe PC server, which is using dotnet or for that matter any JIRA PC server, yeah.

C

So streaming is another key feature of gr PC, so gr PC, utilizes H GB to to enable streaming, and there are different method types in gr BC. So we have a simple unary call. So a client sends our requests on a server, a sense or response. So this is what most people will be familiar with, and http1.

C

We have service streaming, so this is what our code sample before showed: we've send a single request. The server could then send multiple responses, client streaming, which is just a reverse and bi-directional streaming. So a bi-directional stream. The client would establish the connection with the server and then they can independently send messages to each other by that by that TCP connection.

C

Another key feature is what we just talked about now: multiple languages.

C

So there are many implementations of G RPC, both server and client, and these are all organized and developed by the cloud native computing foundation, and there are a common set of interruptus, which all of these implementations follow, including a1, which we're writing to make sure that they can talk to one another.

C

So you're, probably thinking wow G RPC, is great. Let's, let's throw rest and Jason am a dustbin and Levangie of Sealand forever. There are some disadvantages with gr PC, so one is browsers, have great HTTP to support.

C

Unfortunately, JavaScript API is inside the browser haven't caught up, so XML HTTP, request and fetch, which is what you would commonly use and JavaScript to talk to a web server, don't provide a low-level access to http/2 frames and because of that, it's not possible to write a proper gr, PC client in the browser.

C

There is a workaround, so there's a technology called Chi IBC web and it provides limited support for calling Geo pcs services from a browser. So this is developed by the GOP C team and the way it works as there is a proxy which sits in between your geo busy services and your web browser. So the web browser would use the CIBC web client, which can talk either HTTP one or two, and then the proxy would then translate that call into a real actually a GOP C call and then fought it on to her.

C

Chip is G RPC service. So there are some downsides to that. For example, if you use ng RPC web, you don't get bi-directional streaming. Support, I think maybe client streaming as slightly spotty as well.

C

Another disadvantage is it's not human readable compared to HTTP one and JSON, both of which are text formats, um so you require additional, tooling or additional steps to be able to debug Qi RPC calls so, for example, I have Wireshark open here and I'm looking at a TCP packet for HTTP 2, but the tool was able to recognize that it is a Qi, RPC call and I'm able to view some data about the GRP see call in this case.

C

I can see they're calling a greeter service and I say hello method, and this is a response. I mean I can view the view the response down here and you can see some utf-8 tix shows up, but for most case it is not easy to understand. Now there are ways to improve this, so we talked earlier about using Jason rather than protobuf in our messages.

C

So one strategy is to use Jason when you're developing and when you're, developing annual on your local computer or in a test environment and then switch to zone protobuf and production to get the performance benefits.

D

Yeah I would actually stress the part where you know so it's all about to link and because G RPC, you know, is getting a lot of traction with the community, the expectation it's like over time. There's gonna be more and more tooling, so we're hoping that there's gonna, be you know less of a problem in the future because, like if you have like all these tooling, they already have like built-in ER PC support, for example like Wireshark, plugins and stuff. Like that, you know you basically be using. You know, gr pcs.

D

If you were using code Jason, because the tooling would provide you that support.

A

D

A

Got a fun question on twitch, for you question is: is WS HTTP binding gonna be supported, injera pieces.

C

You could write a proxy I.

A

Think maybe the fair thing to say here is like the transport is pluggable. You could do it if you're adventurous, there's nothing first-class for now or for the foreseeable future.

C

A

C

We're focusing on HTTP, 2 and protobuf, so those are the commonly used ways of communicating using G RPC, but we're always opened providing like API is for people to maybe extend it, and if you have ideas like that, create a create an issue on github we should. We should show that github repository at some point and we can. We can talk about it.

C

Okay, this is the last slide. So this is what what are we doing in Donna core 304 G RPC? So we have some existing packages. These white packages are the current currently publish on on nougat, and these are the C core implementation of G RPC, so G RPC core has a native server and Native Client, those rap binaries which come in the nougat package and then Google dot protobuf is the serialization handler for for protobuf and then we're shipping and sex.

C

So already we are shipping GBC, ESP net core server, so that server implementation of G, a B C, it's hosted on kestrel and it's able to integrate with the rest of your asp.net core ecosystem. So you can interrupt with middleware authorization, middleware routing and so on, and then gob c.net, core dot, HTTP client, that's a managed client we're going to be shipping in preview six and that uses HTTP client and it's new HTTP to support for a GC, 2 client. Actually.

E

Four I think this diagram is a little bit out of date, because we have decided to split some of these packages into more packages. For example, the client has been renamed to ERP C, dot, client, and then we were adding other integration support for essentially a client factories and integrating that into the. We also have server-side integration for that as well, and that comes in gr, PC Dyess me net court server that client founder.

A

E

So there are quite a lot of different packages and we to simplify that we will be looking to add a meta package, so we can essentially make it easier for people to reason about how to which packages are actually needed for the workload and on the topic of essentially packaging and what we're releasing. Let me just address a few of the questions I see also on YouTube and twitch, one of which was essentially support for tooling, for example, the protobuf item that was demonstrated in a few of the previous samples.

E

So the support for tooling actually comes from the GRP c. Dot tools mostly comes from your pc tools, package that includes proto C, essentially the protobuf compiler and a lot of the plugs plugins in there, as well as the msbuild integration, and that is already currently available. So essentially, if you're on already, you can use that if you are not looking to run on the kestrel server, so that is currently already available. We are also contributing to that to make it better for 3oh as well.

E

So we will have updates in that area and then there were yeah. So for the tooling side, most of the support comes from ER PC, not tools and that's currently well so I just wanted to address that question. Just.

A

Question James was that your last slide, because there are a couple of questions that have accumulated. This would be a good time to good segue to try and address those yeah.

C

That's that's what okay.

A

All right so, let's see.

E

So another thing that has been brought up in the YouTube chat is so you said that your PC doesn't work very well in browsers because of limitations in support with especially JavaScript. How does that affect something like, for example, blazer? Does it suffer the same limitations? However, certain.

A

The limitation was, you, don't have access to HTTP 2 framing from browser ApS like xhr and fetch blazer doesn't use that blazer uses either web saw like laser negotiates but use WebSockets or it can fall back to. You know sse and things like that. So we don't have this limitation in blazer.

C

I'm not sure like server side, blazer there's, certainly not this limitation, I'm, not sure about client side blazer. So.

A

The the limitation would be, if say, for example, in future, we wanted to support a HTTP to transport in signal or then we couldn't do that, but the way Blazer works today because it uses WebSockets. We don't have that limitation right.

C

Mm I I, don't know, I, don't know how client side blazer does its HTTP requests like if someone creates HTTP client and creates call and so client side blazer? Is it using some sort of don't net or as a using the browser stack front? So.

A

I could be wrong, it's my understanding that it uses the browser stack and we use fetch and a fetch isn't available. We polyfill right.

C

A

D

Okay, just see it a little bit more background on why G RPC protocol wouldn't work well with the browser. Api is that G RPC, like the purple definition uses HTTP, do trailers and that's not exposed to know in the browser API, so you're, basically not able to like return back the status of an RPC, which obviously is a problem. So you know unless you're using those parts of HTTP to spec. You won't have a problem.

E

And another couple of other questions I want to address, one of which is essentially, how does G RPC compare with signal RN, something like socket IO, so I guess for the first portion is that ERP C is a much lighter weight, RPC framework and yes, there's similarities with signal R. For example, there is streaming both sides, both directions and then they're, very similar kind of like RPC, like frame rates, but ERP C is based on code gen.

E

It is designed first and by that I mean you, write a definition filed the protocol, buffers descriptions and then that gets code generated into c-sharp or, whichever other language that you are working with, and then that is essentially the contract. Whereas for a signal, are it's a little bit different in terms of that? It's you. You write the c-sharp code first, so it's called.

C

A major difference as signal I will keep track of all the connections, so with signal our when you connect to a hub that hub will then keep track of that connection, and then additional people who connect to that hub can then do things like say: send a message to everyone else connected to this hub, but me or send a message to all the people in these groups signal GI PC on the server doesn't have have this state so each connection to the server there's, no giant pool of connections that you're able to easily observe and then broadcast out to all of them.

C

It's very as much more similar to we're a regular wrist connection. Where, like you, can't easily observe all the other people who are happen to me, making REST API calls and broadcasts to them. So.

A

I think we're seeing a bunch of questions aren't like signal are versus your PC I think. Maybe we should like big break from on some questions. Actually jump into some chord show some bits we have and look at. Let people see the app model, because you'll notice that some things are just easier to do, especially like browser interactions. The app model in singular lends itself well to that. It's not something you'd necessarily even want to do with your PC, even if you could so I think, let's, let's do that.

A

Let's jump into some code, alright see what your PC actually looks like what we're shipping today what's coming up, and then we can come back to questions cuz I, don't wanna not show off cool bits. Sure okay sounds.

E

Good, let's switch to our demos.

A

Alright, so we've been shipping your PC for about three previews, now I think starting from preview, four or five. So what we do ship is. We have a template that ships in the Box in Visual Studio. So if, if folks haven't seen it before, let me just you know: you want to create a.

E

A sweeter web application right thank.

A

A

Sorry folks, I am, are you all right? Oh yeah sounds like yeah, alright, so just go through with this, and so we have a G RPC service template. This will generate just the server side, cord it's been available since b5, so you can go ahead, get started, I'm, gonna, back out of this menu because I already created a project and I added a bunch of extra stuff to show you folks, but in which.

D

A

Just closed oh did I. Oh my.

D

A

You're right, I did oh, it's. Okay, no worries.

A

Start be right: okay, yep! So let me go ahead and open that project again and.

A

So when you create the project, you get a single, you get a solution with a single project. It's a server-side project Oh by default, we throw in one proto file, so this is similar to what James had shown earlier. This is what a protobuf file looks like. It sits on DSL here's my package name, here's my service definition here are my like actual endpoints.

A

So in my case I, this is a unity call, it's a single required single response pneumatic, and these are my message, definitions, so the flow that you would do is you would make all your edits here. For example, if I had to define another RPC I could go ahead and do that. Let me just change this name, so this like right and then we have core generation, that's also already wired up and what you would see get generated would be. This read a base write. This is the generated code.

A

It's wired up via design-time, build right now, but what you would end up implementing would be this. This would be your actual service definition. I can jump back to the profile. Sorry guy, so.

E

If, for example, I was on older I guess using older bits or your older packages or older tooling, where design time build is not enabled how would I run the code generation, that's what happened so.

A

I think the point John is alluding to is: is the actual code generator that runs that takes your your protobuf file and generates base class that which you then need to implement. We've been working on tooling, so, based on what bits you have you might not have design time bill, so you might have to trigger an explicit build, so I can trigger an explicit build just to show you.

A

It looks like and I'll go to the earth, and you know click on this show all files neat little thing that I never really use, but we got a degenerative cord. You can see. Oh it's.

E

A bit harder to see it, can you see oh.

A

Yeah, let's see so.

A

This is really tricky with my trackpad, but I think we can do it. So you can see all these all this code was generated as part of the build and then you'd go ahead and implement something like this right. So in this example, where I went ahead and added another service, if I go to act, my actual service definition right so I have I want to like now implement my new method. I can just go in here. Override and I.

A

Have my new RPC that I can go ahead and implement just Oh question? Is my font size? Okay? Yes, you know.

E

It's I think it may be a little bit small if you can make it slightly bigger. That would be all.

A

E

I make it fine. Okay,.

A

Okay, so so what I've done here is it's a couple of things I wanted to show off. Obviously, is. This is recent as a visual studio, I think 16-1, where we have syntax highlighting for writable files? That's new go ahead and try it out. Folks, if you run into any corner cases, go ahead, leave us some feedback via visual studio. We look into it, but so far it's you know been after one release. It's looking pretty.

E

Good and we also have a little bit more integration in terms of syntax errors, for example in the probe of I, also.

A

Have stuff like that, that's a fun thing to show: John: okay, let's see how do I mess this up? Okay, let me take a message that doesn't exist and I go ahead and build.

A

B

E

Your validating.

B

Against the great.

A

B

Above definition,.

E

So we have the errors linked back to where they occurred in the proto buffer, so it'll be easier to essentially find out where all your errors are so I like this flow John.

A

I think we should keep.

E

It this way I keep doing stuff. You tell me what else yeah sure well.

B

One question from Tim Tim here is asking you'd like to see a direct comparison. Maybe this is just overall feedback like to see a direct comparison between Web API G RPC for SP net dev. So, like obviously, you've pointed out the perf benefits, but what's the difference in dev in the develop development for experience, yep.

E

That's a great idea and I: we have a little bit of documentation on this. Essentially, our current documentation compares against rest, for example, but we can share the links for those tonight. I've gelled I'll, look into that. Well, I should continue with a double yeah.

B

And one other thing just well: I'm rudely interrupting here, Michael Powell saying he thinks he thinks that the code generated should yield partial classes, especially for services, so that we can, you know, extend them without having our code over it and when you generate and.

E

I generate the code, yes, the generator classes are partial, actually, okay,.

A

E

A

Actually jump into that real, quick, so yeah, this is the generator class, is a partial class.

C

E

C

Why you would want to extend it using partial because you would always want to inherit from the base class, because if we look at our greeter base at the moment, the virtual implementation of say hello would do nothing more than throw a a, not implemented exception, oh yeah and unemployment at status code. So you always need to inherit and override these method. Stubs and these these method stubs are useful, because the the visual studio tooling will just automatically do code completion for you and every everything strongly typed.

C

That's one of the benefits and my opinion of GRB see the that cogeneration will handle, generating your messages. It'll handle generating your based service, your client, so you don't have that overhead of first creating a service and creating the messages then having to copy and paste or maybe to placate duplicate, duplicate, creating the messages I'm, a client then creating a client that strongly typed and building up URLs and doing serialization, all that is handled for you and G RPC, and it provides a great end-to-end experience.

C

All you need to do is just share these proto bar files between your projects or, if you really don't like the idea of sharing protocol files, you could even use code generation once in a shared service, have a target, it's standard, Oh, compile it and then shared share that compiled NuGet package between your applications. So there are many different ways of doing it and I think it's a really great benefit of QV C. If we're coming back to that earlier question about the development experience, yeah.

B

Speaking of that, there's a question from twitch that I think I answered, but I'm not sure it's the best answer. So the question is: when I'm developing with rest I can use postman or whatever to validate things when I'm developing against G RPC I need a test client and my what I think is. The answer is that you're developing that for me, you're, generating a test client for me that I can use. Is that kind of the the main thing or should I use some something else to do test generation? There.

C

Is a tool like postman I can't remember its name off the top of my head, but it is a GUI tool where you can add hoc query key RPC services, ad-hoc, query, ng RPC services is going to be slightly more difficult because the proto bar file contains the definition of your message. So your first step will probably be loading. The proto bar file into your ad hoc client and then making calls.

C

um But let me let me look it up, because I have I haven't seen someone talk about it, but I can't remember the name off the top of my head and.

A

Also and I think Yan can actually maybe speak to this. There is a CLI tool that you can use to as a client, I think and I think believe. There's a canonical mapping between like text to how you do a representation of proto. So if you really need to like hack something quick and dirty, there is a command-line tool that is available as part of the GPC Reaper cool.

D

Do you want to add anything that young, oh no I, think you described that pretty? Well, yes, so there is a command-line tool for protobuf there's, actually a canonical textual representation and also with protocol buffers version 3. Oh there's a canonical mapping between Jason and protocol buffer messages, so you can do basically like transformation one way or the other, and you know it's going to be cleared the way the mapping works. So.

C

The toll I was thinking of it. It's bloom RPC, so it's open source. It's on on github I, just shared the link and YouTube, but it's a it's a GUI tool around around querying protobuf services. Well, RPC services.

E

Okay, maybe we should continue with the demo, looks like or okay I guess we can take a few more questions. Yeah.

A

It looks like I love lost Network for some reason, I see I. Think while I try to get this back John, you can show the command line tooling experience that we've been working on. Why don't we switch over to your machine sure? Let me throw you right onto the bus. Let's.

E

A

E

That to show up I think this cable is actually just cursed. So, okay.

A

I'll give it back to me yeah. We can like physically switch over cables me okay, so the thing that I was going to show off is so I went ahead and added another profile. We have I John. We shipped item templates for this. If.

D

A

A new service you can use the item templates. It's part of the CL line. Now we're gonna add it to visual studio soon.

A

So we go back so I went ahead and added a new, so create a new file right. You know described my service describe my messages, but then there's the point of actually adding it to my project. So I know what code generator to set up. So this is this is fresh bits. I literally went over to the visual studio folks copied some DLL so from that computer and showed up here for this demo. So we have this connected services node in the solution Explorer that you're used to seeing for projects.

A

This is what we use for connected services when you connect the authorized your services, we have this node that we've introduced called service references. So if you go ahead and see like I actually have some stuff listed under the gr, PC node there's also the open, API stuff. That I think we can do a future separate talk on, but it's going to be a very similar experience, so you can go ahead.

A

Click this or this allows you to either point to your PC to an existing product, I'll so or there's also the the URL based thing and since I've already added it I can click on edit. The other thing you can do is you can specify options to the generator on whether you wanna generate just a client, the service tab, both client and the server or nothing at all. You just want messages from that.

A

So this this UI that I'm showing you is still hot off the press, it's supposed to land in 16-2- and you know- hopefully this will make you more productive, but you're going to add this. The other thing you can also do is shoot or increases font size. I think this might be able.

A

So we have this tool. Is it that Ned your PC John? Can you speak to this I.

E

Don't know it's dodgy our PC and then you can see the help text. So, with this tool, we can manage essentially jar PC references, so you can add, for example, reference to a local protobuf file on disk, also remote links to put above files via a URL.

E

You can also refresh any existing references that are based on URLs, as well as remove references, so we've added that support. You can configure whether again, similar to the vs tooling, you can configure whether it is for server or for client for both a couple of options there as well and again. These tooling gestures will also ensure that, if you, for example, specify a server service, then the necessary packages, the NuGet packages that are that contains those api's will be added to your project, is essentially automatically yes.

A

And that's a good point that John mentions, because, especially if you're using a client, it could be like in this example, I have a client which is nothing more than a console, app you're not going to have these package references that you need. But when you go to that dialog, let me go over here again connected service and you add your first repeat like proto document. It will go ahead and add those package references for you.

A

So if you don't know what you need to do to get started, it's gonna be super straightforward, just go ahead and add that in and they'll bring in all the dependencies- and you have a strongly typed client available that you can just already consume.

A

B

Slightly snarky comment here, but I think it's a good question is when you bring up, add service reference and you see WCF there and then there's a is this. How does this compare to WCF? Is this just the new shiny thing or what's what's the big kind of differences between this and WCF.

A

So hope you want to take that the the easy part for me to answer is there is no WCF server our net core 3. So if you want to do if you're looking to move away from dotnet little more thing in WCF and you need some solution in net go or three, this is probably what you would do and.

B

Reason for that at its WCF is bill is very window specific and built on windows internal stuff, whereas gr pc is cross-platform cross. Everything right.

A

I the joke I like to make. Is we had this thing in Microsoft where we had not invented here syndrome, but we wouldn't use anything if it wasn't built by us I like to say what is the new Microsoft where we used the proudly found elsewhere? So we saw a very popular RPC framework. We saw people were using it in the community, so we, like you, know, let's get behind where the community momentum is right. So there are, there are yes, there are functional differences, we're looking at what we can do.

A

If you know there's a you know, migration path and to that end, I think you know I, don't know if you've covered this in one of your community links, John think mark Rendell did a you started doing a blog post series on migrating from WCF to G RPC in that net. Core I think that next week, okay.

B

A

There will be something function, unfortunately, I don't have the best of answers for you like there are things that, were you know, still finding out and I I, don't know I I, don't think me personally can advocate for this app model switch already. I think it's case by case, and you know, will improve the guidance as we go along I think the link that John shared also did some compare and contrast to it right.

E

Right, yeah, yeah, John I actually sent the link to you. So if you can post it in to twitch and YouTube, that would be nice I think.

C

Gip see it improves on wucf, see if and and some key ways. I've worked on a number of real-world projects using WCF and there are pain points with using it that I think a PC is solved.

C

So streaming is a lot easier, especially bi-directional streaming.

C

There's, obviously not enormous amounts of XML documentation to wire up your your address, your binding and your contract, your your ABC and WCF, and what's WCF, to generate a service proxy. You need your server running because it's generated at runtime well, GRP see the generation happens from some proto files, which are just checked into source control, so those proto buy files, your source of truth, rather than going through the hoops of generating a creating a service contracts and then your messages and then running your server, which generates winds tool which then generates your service proxy.

C

You can instead just generate the service proxy directly off the proto files, I.

E

Think we should probably.

A

Go continue with.

E

A

A good segue because you know James alluded the fact that you know you have a single profile, that's a source of truth. So if you actually see I'm in my client project right now, and if you see the files that I've added the location on disk, but these are actually in the server project- I, don't know if it's visible and zoom in. But if you see it's actually walking up a directory and it's in the server project versus if I look at oh shoot versus.

A

If I look at the server project, you'd see it should be right here you see the difference so yeah. It's like your profiles, your source of truth, so I've gone ahead and added that and let's do you. Wanna jump in the client cord, John yeah.

E

Let's take a look at the client side: okay,.

A

So let me just quick I've been showing the service, but I haven't actually spoken to what it does. So it's a pretty straightforward service. It is a client streaming service, so I have my.

A

It takes a stream of ad requests at request is a simple message that just has the value and the it returns an ad reply, which is just the sum of everything: you've stream. Do it. So let's go and look at the added service, real quick, it doesn't do anything fancy. You know just a request. It reads the current value in the request stream when we reach the end of the require stream. It just replies with the song. Let's go look at the client now: oh okay,.

E

So why are there two different kinds of clients, okay,.

A

So like we do before there was an existing GPC implementation and, up until I'd, say a couple of days ago, that was the only client implementation available or darknet. We have recently, if you guys been watching, you know core facts: repos we added HTTP to support and used to be client. That was one of the last stumbling blocks to enable a fully managed to your PC client. So what I have up here is what the old pattern looks like.

A

There are a couple of things that we're still, you know trying to figure out what the shape of this API is. What is the analog to channel, which is a core concept in your PC Cole? What the analog to that is in the managed implementation, so this API is subject to change, but it's still super straightforward to go ahead and use it so, and the good thing is like- and this is also applicable for people who already have stuff and dotnet using your PC core. There is the migration.

A

There is a part to move to the dotnet core implementation, as you can see like in this case, I can pretty much straight forward. I'm, just gonna. The actual invocation of the are the client streaming call is just a streaming at async and I'm, passing it a client which I'm switching I'm gonna not use this. For now, I'm gonna use a manage client version, and so there's nothing fancy going on here in terms of creating the client.

A

You know it's just a new ways to be client at the base address, and then we have a static helper that hangs off of your PC. That will create you an advert, client right and then I actually invoked this. So this example is pretty straightforward. All I'm doing is reading numbers from the console up until I get an empty line, which is the end, and then I would once I do that.

A

I output, the response I get from the server as to the console, so I go ahead and run this whoops I'm, starting from the command line. There's no reason I particularly need to do that, but.

A

All right so very fine and let's see, watch logs, okay, so Brian connected, so I can send a bunch of stuff, so I can send a bunch of so this is live streaming as I type. These in and I hit return these get sent over and when I'm ready to finish, I got the Sun back from.

E

So what I think we also want to demonstrate some do a quick demo of the interceptors okay. So what what are interceptors on like? Well, there are a gr PT concept there. If you are familiar with SP nut core concepts, it's very similar to middleware. So essentially, when you process AG RPC call, you can add these interceptors that will do pre-processing or post-processing on your request or on your requests on the way in or on the response on the way out.

E

So we have gotten some community feedback asking that for interceptor support, because there's already a lot of like libraries or other people have started using interceptor to solve a cross-cutting concerns, for example retries or that would be a client one, but on a server-side interceptors can be used to handle timeouts, for example, for all of your calls and make sure that no of your JRPG calls go over a certain time limit. So we have yeah I'd like.

A

To jump in, if there's anything that you've seen or you heard because he's been part of the juror PC community way longer than any of us had.

D

Yeah, to be honest, I was reading through the questions, so I wasn't really.

A

Like in question, what have you seen as usage of interceptors in the Jarod PC community? Oh.

D

Well, so for authentication, or also like collecting stats.

E

Logging and those kinds of things, probably like potentially error, handling and those kind of concerns as well and I know. This question will probably come up, which is, if I'm, using a yes, you know if I'm writing an ASP Nootka rap. What are the when will I use, for example, an interceptor versus a middleware, because the concepts are somewhat similar. The difference between interceptors and middlewares is that the middlewares will operate on the HDTV messages and and they will not have been deserialized. So all of your messages are raw messages.

E

You have the access to all of the things on HP contacts like headers and raw byte stream, but if you want to operate on the actual G RPC messages or that your PC types, then you will have to do that inside the interceptor pipeline that we have. So if you want to operate on raw ACP DHCP, then use middlewares. If you want to operate on your PC messages, use interceptors, okay,.

A

So I'm going to show you folks an example, so John's written this Mac streaming request time or intercepted John. Do you want to speak about what it does and I'll go wire it up, and we can then show how it works and then jump back into how it was written. Sure.

E

The concept here is that, essentially, if you have a client streaming interceptor, we want to make sure that the client stream isn't open for too long, so this interceptor will start at time. Well, it.

D

E

Essentially, cut off your calls after a certain time has elapsed and you can configure that from inside a Interceptor and right now the default is 10 seconds, but I mean for our demo. We would probably make it shorter, but yeah.

A

I should make it short.

E

Yeah and but I think after just demo we should probably start wrapping up because okay, so the.

A

E

Thing we demo so.

A

I got so winding up. Interceptor is pretty straightforward. You know we have the interceptors that hang off of your PC service options and I've gone ahead and added this, and this interceptor takes the time span. So I go ahead and woops. Sorry, let me go ahead and run this again. So in this example,.

A

So have my planning and now my client connects and it pretty much terminates the client stream in 3 seconds. So if I can I can try and send some stuff. But if I go over that limit.

A

And yeah, so my client stream drive terminates if I try to do something else, I get a RPC exception, so you can. You can see how people can use interceptors to enforce policy things and John already alluded to like timeouts deadlines. Yeah things things like that- and this is one of the pieces of like overwhelming feedback that we got because our you know, initial prototype didn't happen and we went to jail, PC cons and we saw everyone's like now. That's the one thing we have to have. We have libraries in our organization.

A

We do policy based enforcement using interceptors, so we've recently added support for that as something we're. Also looking for. You know, feedback on the area if.

E

There's any other er PC features that are definitely needed and give us that feedback and we'll try to add those in.

A

Right, so that's that's all we have for demos Meakin. There.

E

Was another question that was brought up a little bit earlier on YouTube and I'd like to address that which is right now gr PC is using a CB 2 and we're demonstrating it is essentially going over to network and stuff. Is there anything that we can do for in process or a cross process, essentially not going over the network? So the way our server stack is built is.

A

Our transport is pluggable, alright and so for your PC server implementations, like the other ones in the community, like the natty based one, has a plug a bullet believe it has an in-memory transport. It also has like a UNIX domain socket base transport, so it is doable. We have not looked into it yet there's. Definitely some like area future investigation. The tricky bit for us would be our HTTP client stack, does not support a pluggable transport, so we don't have a good answer there.

A

Yet we will look into it, but getting a pluggable transport on the client side is is vaporware right now it's a cool idea in our heads. We've done zero work towards achieving it all right.

E

I think that is all the time we have so I think. That's probably a good point for us to finish. Yeah.

B

If you can switch back to the the camera view and then for folks that want to kind of keep up when it you, you know you mentioned several times wanting to you know, have feedback and all that where's the best place for people to go for that. It's at just file an issue on that SP net repos or we.

E

Actually have a repo under the gr, PC trpc2.

D

E

So the actual link would be github.com /g, RPC /g RPC net. That is where all of this work is currently being done. So if you file an issue there, we will see it and we will respond to it. You.

A

Know on Twitter and me and James around and I can poke John. If you ping me on Twitter, so we can get all your questions answered very.

B

Cool all right won't have the the link to that in the show notes, and thanks a bunch. Everybody thanks for your time and special shout out to yon for stand up to very, very late for us and Munich much appreciate it. Thank.

A

You yawn thanks folks John. Do you know how to end this.