This is the first I've heard of it. So I can't say much about the project itself. However, by the look of it DeskGap is at least taking the right approach, where as most these other projects (there seems to be a new one every week) miss the mark.
Electron Pros:
1. Consistent JS runtime/API
Electron Cons:
2. Huge distributions (bundling a full browser)
3. No re-use of components (every app is huge)
Electron Pro?/Con?:
4. Consistent rendering / WebKit monoculture
By the look of it DeskGap keeps Electron's biggest pro and solves its two biggest cons. The only "issue" is that there may be rendering inconsistencies. However, given we want to avoid a WebKit monoculture, I see this as a pro more-so than a con.
With that being said Microsoft have themselves migrated to WebKit; so the WebKit monoculture looks likely regardless. For consistency I guess that's even another tick in DeskGap's corner, although it'd be nice if on Linux (for example) it could use Firefox when present.
Carlo is interesting for a similar reason, but perpetuates the WebKit monoculture.
zserge’s webview has been adopted by a bunch of maintainers who made Rust bindings for it. They have forked webview to continue maintaining it, and are now gradually rewriting it in Rust.
Made on top of that is Tauri, a cross-platform webview-based app just like WebWindow. The key differentiator being that they are actually committed to continuous development as they’re trying to get established as an open source company.
We're doing it. It's (misleadingly) called server side Blazor. The UI is seamlessly and naturally web, while 95% of the logic is plain old C# running on the plain old core CLR.
Apologies, I meant to state that Windows WebView uses MSHTML which is IE10/11. I see in BETA right now is a Win32 API for WebView2 which uses Edge(Chromium)[0]. Not sure if you edited your comment for clarification or if I replied to the wrong post as my initial comment no longer makes sense.
This looks amazing - can you please add a license file so others can actually use it? (Hoping for MIT or if not then Apache 2.0 or BSD or whatever reflects your intent and preferences)
I'm potentially the only one being skeptical. Don't get me wrong, it's nicely done (for an experiment), but we also did similar experiments several years ago.
At the fundamental level such an approach works, but it brings several issues that are not addressed with it (in contrast to, e.g., Electron). The most pressing one being "I don't want any browser - I want a browser that I know is capable of doing what I want".
So most of these techniques used to have the problem that on Windows "any browser" was always "Internet Explorer"; but while I don't disagree that this is still an issue, it is much much much less of an issue now that Internet Explorer is Edge and Edge is Chromium (which I am not saying is a good state of affairs for the web as a whole, but makes this use case not fall as easily to this complaint).
But I don't think Edgium is replacing Edge/IE as the system web component. It's no different than having Chrome installed except it replaced the Edge icon in the start menu instead of making a new one.
I'm skeptical for a similar reason. This isn't really "crossplatform" it is "capable of running on a very small list of platforms which bundle a web browsing component."
This looks like a nice solution to save some time for cross-platform developers on .NET Core. I'm sure we're going to see a lot more of this once the new Chromium version of Edge is released mid Jan 2020 (which will include the new webview2 component).
If you don't want to add an external library to do this, or your source is already platform dependent, just use the default web view component of each operating system (WebBrowser for Windows, WKWebView for macOS and WebKitGTK+2 for Linux) like this library does.
To work around some of the ugliness of the old IE rendering engine used in the current WebBrowser control on Windows, just add this meta tag to the HTML file to ensure it uses the new versions of IE to render the UI: <meta http-equiv="x-ua-compatible" content="ie=edge">
You can disable most non-native app behavior in JS (like scrolling or the default right click context menu using javascript). You're then free to use whatever front end framework you want and pass data between the web-app and native code using JSON.
I think setting x-ua-compatible ie=edge doesn't mean use the Edge EdgeHTML engine - it means enabling "Edge mode" in IE, which means "the latest available version of IE" and you end up with IE10 or IE11.
EdgeHTML can be embedded, but only using the WinRT API, not the ActiveX WebBrowser control.
Welcome to the web, where browsers are named after existing browser concepts in as confusing a way as they can come up with. “You know how browsers have chrome around the web content, and how that’s the accepted term of art? Let’s call our browser Chrome!” “You know how we can tell IE to run in edge mode? Let’s call our replacement browser, which IE compatibility mode won’t be able to use, Edge!”
While this definitely sounds like a good alternative to everything bundling its own Electron, I do have some questions.
To make sure I understand how viable this solution is...
> On Windows, WebWindow uses the new Chromium-based Edge via webview2, assuming you have that browser installed (it could fall back on older Edge if you don’t, but I haven’t implemented that)
So this will currently only work for those running the absolutely newest version of Windows 10?
(But if the user does, you get a reasonably up to date and capable web-engine, which sounds good)
> On Mac, it uses the OS’s built-in WKWebView, which is the same technology behind Safari
Sounds decent, I guess?
> On Linux, it uses WebKitGTK+2, which is yet again a WebKit-based technology
Isn't WebKitGTK horribly outdated, or have things improved recently? Couldn't that be a source of issues down the road?
I would recommend starting with the server-side variant of Blazor on a non-production test app to familiarize yourself. You can easily move to client-side WASM mode once you need to (or feel comfortable doing so) without any major changes to your codebase.
Our long term plans (aka by end of 2020) are to rewrite all of our internal tooling from SPA (RiotJS) w/ JSON APIs to server-side Blazor. The principal motivations here are productivity gains. It is a LOT faster to write a razor view in C# with direct invocation of your business logic from view logic, than it is to determine what a clean JSON API contract looks like, write all the boilerplate for controllers, and then play the back-and-forth game between chrome devtools and visual studio with 2 separate debugging contexts. After you write your first LINQ .Where().Select() to populate a table in Blazor, you will truly understand what I am getting at in terms of productivity if you have ever tried to do the same sort of filter->map stuff in JS.
With Blazor, I only need to know C#, HTML, CSS and just enough JS to write a blazorhelper.js interop which reads and writes my session cookies. I set 1 breakpoint in visual studio and it gets hit when I press some button in the browser.
With SPA, I have to be at the top of my JS game, and also pull some SPA framework into my mental context which must then be shared with all of the other C# business logic in scope. I have to manage breakpoints in the browser as well as in visual studio. I have to deal with 10-100x the amount of JS code.
Additionally, we don't even want to move to client-side variants of Blazor for our use case. First request latency and security are the 2 largest motivators for us here. Server-side exposes precisely zero things to the client you didn't want to, as opposed to SPA which makes even unauthenticated clients download your entire application. Also, "the server" is substantially faster these days (even more so going into 2020 on AMD's warpath). For the number of concurrent clients we are anticipating, server-side Blazor is actually our best option in terms of UX.
I would be curious what aspect of Blazor feels immature to you. It is certainly a new(ish) thing, but the fundamental aspects of it (especially server-side mode) have been around for decades and microsoft has leveraged a lot of existing or otherwise proven ideas here. I would agree on the client-side aspect being "immature" relative to server-side, mostly due to WASM and the need to push a very large payload to the client upon first request.
I absolutely see the benefits of using Blazor. But it feels immature because I see a lot of tutorials with ways to work around little bugs and problems.
But orcourse those tutorials could be from before the official launch which was just some months ago.
I think Blazor looks a little like Phoenix Live View. And imho this is the way to go. It is way more productive for the reasons you describe as well.
No but I agree, it seems a little risky. I would wait for the next generation of web assembly changes to come down the pipe before investing in this too heavily.
The blog specifically states "this does not involve WebAssembly", so it seems like this uses Blazor Server and not Blazor WebAssembly, meaning that it doesn't run any .NET code in the browser space at all. While this is also very new, it's an officially supported part of ASP.NET Core 3.0. I'd probably not use Blazor Server in a web app just yet since it requires an active connection to the web server, but this is all local so I'm pretty interested in trying it out.
Of course this is still an experiment so it's very risky to use it, but it's cool and shows promise.
Edit: Looks like maybe this thread is talking about using Blazor at all instead of within the context of the post, but I what I said still applies so I'll leave it up.
My biggest problem is that WKWebview on Mac just seems so slow compared to Chrome. I have a canvas in my application and it just doesn't seem to render as efficiently as it should or with ease as it does in Chrome. Does anyone experience similar issues?
It might have to do with the retina settings that rescale after rendering. I had this issue when we were testing 3DC.io on some Macbooks. Play around with the scale and css a bit, there is some information on Stack Overflow if you search for canvas+retina.
Edit: Also check out devicePixelRatio, that might be causing the issue as well.
Ohh! That actually was it. My canvas was a very large size - like 1900x1200 and Safari drastically slowed down - I readjusted for it to only take page height and width and it works nicely again. Thanks!!
I've never experienced this. It sounds like you are hitting a bug, but more investigation would be needed to substantiate whether there is a real difference. Have you tried to make a test case? Please file a bug at feedbackassistant.apple.com with repro steps!
this is really interesting! i'm wondering if there's a variant of this project out there that still uses node/javascript, but similarly avoids the chromium packaging and leans on the OS's rendering engine.
1. Webview: https://github.com/zserge/webview
2. Lorca (Chrome): https://github.com/zserge/lorca
3. Carlo (js/Chrome): https://github.com/GoogleChromeLabs/carlo