First, it's now called Rust for Windows (or windows-rs), because the scope has expanded to include (fairly low-level) bindings for all the Win32 and COM APIs in the Windows SDK.
I strongly object to calling it a toy just because it doesn't have certain RAD conveniences. I'm happily using that crate (though currently only the Win32/COM portion) in the native parts of a Windows app I'm working on [1]. Yes, I have to use a lot of unsafe when working with the Win32 and COM APIs, but I keep my unsafe blocks as small as possible, and try to do as much as possible using safe Rust abstractions like slices and the various string types.
FWIW, windows-rs recently removed the bindings for the Windows.UI.Xaml namespace, on the grounds that they were nearly unusable, and because the in-box Windows.UI.Xaml is no longer Microsoft's recommendation for app developers. I think the latter policy is a mistake, because some of us actually want to produce something with a minimal memory footprint, using a GUI toolkit that's already in the working set on all Windows machines. But maybe someone can bring back Windows.UI.Xaml bindings in a separate crate.
> they successfully managed to kill C++/CX
IMO C++/CX was a mistake. It was an attempt to make C++ into something it was not. From what I've read [2], C++/CX spectacularly failed to uphold C++'s zero-overhead principle, and the code size of newer shell components in Windows 10+ suffered as a result. Yes, I felt some of the frustration of working with the current (or at least, mid-2020) tooling for C++/WinRT when I was mentoring an intern on the Windows accessibility team at Microsoft (which I left later that year). But I think the Windows developer platform team had to do what was right for runtime efficiency. Developers who value rapid development over runtime efficiency can still use C#.
> the glory days of Visual C++ 6.0 with ATL 3.0
They were glorious, for code footprint. But like you, I wouldn't want to write new code using ATL (and on the Windows accessibility team at Microsoft, we avoided that as much as practical). Luckily, C++/WinRT is a big improvement over ATL, thanks to modern C++. It's just a step backward from C++/CX for some RAD conveniences that were really a better fit for a language like C#.
And to bring it back to Rust, I think Rust's macro system allows for some powerful high-level tooling on top of the WinRT bindings. For example, I think we could have a macro-based DSL for succinctly declaring a tree of UI elements, and that syntax could be more integrated with the rest of the language than the markup/code dichotomy that was introduced with WPF. Such tooling could also avoid the trap of boilerplate-heavy code generation that's typical of the Visual Studio wizards, and be available to developers who don't use VS.
[1]: The non-native part is using Electron. I'd love to write the whole thing in Rust, but Electron in general, and access to Chromium's full WebRTC stack in particular, is just too convenient.
A wall of text, ignoring that C# developers do have to use C++/WinRT, as not everything is exposed. I guess WinDev is too busy writing IDLs with notepad like tooling.
It is a toy when PAYING customers have to deal with that crap tooling versus C++/CX.
And all the Rust stuff is not "hard at work" as suggested by OP, when it is a small subset of Visual Studio has in the box for .NET languages and C++.
First, it's now called Rust for Windows (or windows-rs), because the scope has expanded to include (fairly low-level) bindings for all the Win32 and COM APIs in the Windows SDK.
I strongly object to calling it a toy just because it doesn't have certain RAD conveniences. I'm happily using that crate (though currently only the Win32/COM portion) in the native parts of a Windows app I'm working on [1]. Yes, I have to use a lot of unsafe when working with the Win32 and COM APIs, but I keep my unsafe blocks as small as possible, and try to do as much as possible using safe Rust abstractions like slices and the various string types.
FWIW, windows-rs recently removed the bindings for the Windows.UI.Xaml namespace, on the grounds that they were nearly unusable, and because the in-box Windows.UI.Xaml is no longer Microsoft's recommendation for app developers. I think the latter policy is a mistake, because some of us actually want to produce something with a minimal memory footprint, using a GUI toolkit that's already in the working set on all Windows machines. But maybe someone can bring back Windows.UI.Xaml bindings in a separate crate.
> they successfully managed to kill C++/CX
IMO C++/CX was a mistake. It was an attempt to make C++ into something it was not. From what I've read [2], C++/CX spectacularly failed to uphold C++'s zero-overhead principle, and the code size of newer shell components in Windows 10+ suffered as a result. Yes, I felt some of the frustration of working with the current (or at least, mid-2020) tooling for C++/WinRT when I was mentoring an intern on the Windows accessibility team at Microsoft (which I left later that year). But I think the Windows developer platform team had to do what was right for runtime efficiency. Developers who value rapid development over runtime efficiency can still use C#.
> the glory days of Visual C++ 6.0 with ATL 3.0
They were glorious, for code footprint. But like you, I wouldn't want to write new code using ATL (and on the Windows accessibility team at Microsoft, we avoided that as much as practical). Luckily, C++/WinRT is a big improvement over ATL, thanks to modern C++. It's just a step backward from C++/CX for some RAD conveniences that were really a better fit for a language like C#.
And to bring it back to Rust, I think Rust's macro system allows for some powerful high-level tooling on top of the WinRT bindings. For example, I think we could have a macro-based DSL for succinctly declaring a tree of UI elements, and that syntax could be more integrated with the rest of the language than the markup/code dichotomy that was introduced with WPF. Such tooling could also avoid the trap of boilerplate-heavy code generation that's typical of the Visual Studio wizards, and be available to developers who don't use VS.
[1]: The non-native part is using Electron. I'd love to write the whole thing in Rust, but Electron in general, and access to Chromium's full WebRTC stack in particular, is just too convenient.
[2]: https://devblogs.microsoft.com/oldnewthing/20220606-00/?p=10...