Hacker News Reader

Comment by alkonaut | Hacker News Reader

Modern js frameworks makes mvvm/similar quite pleasant. Some of the binding in WPF is quite complex, especially since you don't get proper type safety for all that effort (there is a lot of stringly typed things sticking out).

On the other hand, the layout model of WPF is a lot easier than mastering html+css.

JS desktop apps still mostly don't work without embedding a browser like electron and that ranges from annoying (e.g. for size/memory reasons) to impossible (for things like a plugin where you have a constrained environment, need to launch instantly etc).

The middle ground is desktop js using system browsers, but it's pretty painful to do too.

View on HN

> On the other hand, the layout model of WPF is a lot easier than mastering html+css

Thank you. I've never been able to figure out if web layout is as terrible as it seems, or if I'm just missing something.

Once you spend the 20 minutes to learn the WPF layout algorithm, you can reason about why every element is located where it is, down to the pixel. To get there with HTML/CSS, I feel like I would have to memorize and recall a huge list of ad-hoc rules, and their combinatorial interactions. Usually I just give up and accept the mysterious 3-pixel gaps that seem to crop up randomly, and I don't even attempt complex layouts.

WPF never quite made the leap from data-binding and MVVM to a component-based + state-management architecture like React&friends. A consequence of that is a hairy overcomplicated mess of a binding system that's full of footguns. But it was still ahead of its time - that one last step would have made it amazing.

They could easily have done this but instead chose to make several new frameworks that didn’t improve anything. Very sad. Now MS has several mediocre not well maintained frameworks instead of a really good one. The constant churn also killed off a lot of open source efforts

This isn't my area of expertise. What are some of the other MS frameworks or OSS projects you refer to?

WPF started XAML, then came Silverlight, WinRT and UWP. They are superficially similar but under the hood the code is quite different and each has its own set of weird limitations so you can't just reuse code. Also, in the case of WinRT and I think also UWP the new framework had less features than its predecessor so in a sense it was worse and less powerful which is not a good way to convince people to port their code.

When WPF came out there were quite a few very cool open source projects that built on top of it because it's really easy to extend WPF and build components for it. But then every two or three years MS announced a new framework so most of these efforts died out. When you look at commercial third party suppliers like Developer Express they have the same component in three or four flavors. this must be a huge and expensive effort to maintain.

"WinRT and UWP" in this context are just two names for the same thing, though

Sort of. They were released some years from each other. WinRT was super weak and never took off and then suddenly UWP came up.

APIs were the same, just with some upgraded and new stuff

I think the various name changes have been extremely confusing and created the perception of more "resets" than actually happened at the code and API level. They dropped the ball by not bothering to have a good compatibility and migration story across WPF->Silverlight->WinUI, but since Windows 8 they've been pretty good about maintaining continuity for WinRT APIs

After trying WinRT and seeing that everything was async (WTF?) and a ton of APIs were missing I admit I stopped bothering to look any further. UWP (the UI part) is also not really a step up from WPF either.

Because any UI code that requires performance should be async and lenghty work must be done outside the main thread.

So Microsoft learned how lazy devs are and how ignoring such advises contributed to bad perception of GUI apps, e.g. Swing, and forced everyone to code the right way.

It is relatively easy to turn async code into sync, if really required.

If that was really MS's reasoning I find it risible. There are many perfectly good reasons to have synchronous IO APIs. Yes, in the wrong hands they can make for janky UIs, but then the problem is not the tool, the problem is the developer.

MS's market was full of amateur-hour developers because their platform (WP7) was weak and lagging behind, thus incapable of attracting the top tier of devs. And then they poured gasoline on that fire with a barrage of silly incentives that were ignored by pros but attracted even more beginners. Now, to put a cherry on that fine mess, they attempted to remedy their bad business decisions with a technical solution that damaged the one asset they really had - an advanced and ahead-of-the curve app-development story. One own goal after another, really.

"Keeping apps fast and fluid with asynchrony in the Windows Runtime"

https://blogs.msdn.microsoft.com/windowsappdev/2012/03/20/ke...

As for the rest, I only agree it was a big mistake not to have a proper migration path.

However, WinRT was anyway the result of Synfosky's team (aka WinDev) being on the winning side, after they won over Longhorn, so not so much .NET love was going on back then.

And we had to wait for iteractive releases of Windows 10 slowly fix all those mistakes.

There are other ways to deal with this besides async. It would have been fine if they had added async but you can’t just remove the synchronous functions and expect people to port existing apps with perfectly fine multithreaded code to async. This would be a lot of work without any payoff.

Whatever their reasoning was it was just plain dumb, heavy handed and unrealistic.

Those feel like they were created by orders from top management, because of some strategic decision (microsoft needed a flash alternative, or a Windows 10 store UI framework etc).

They don't feel like the product that would come if DevDiv did their best to produce a sane UI framework. They feel rushed, semi-overlapping, half-finished.

I imagine it should be possible to make just one more xaml framework which is cross platform, and much better than WPF (and it's derivatives), and much better than the js alternative (Because of deployment, not least).

I'm not sure if there is any incentive for ms to put their weight behind it however.

Hopefully that last xaml framework is avalonia, it's definitely gaining steam.

Yes. But for one of these to really take off, I think what's needed is for one of the giants to throw their weights behind it. Basically "We're betting x billions on it, so you can be sure that lots of companies will be using this tech in the decade to come".

The XAML in Avalonia is great. There is even support for XAML to IL compilation during build so you don't have any parsing overhead when running your app.

> binding system that's full of footguns

Please elaborate?

The UWP XAML flavor now has x:bind which is probably exactly what you are asking for (control bind directly to their code-behind, no DataContext and DataContext inheritance). I've heard rumors about WPF x:bind support.

> Some of the binding in WPF is quite complex, especially since you don't get proper type safety for all that effort (there is a lot of stringly typed things sticking out).

You can add a design time view model to get compile time warnings(and autocomplete) about incorrect bindings.

Reply on news.ycombinator.com