Nobody's doing UWP
posted under category: dotnet on November 14, 2017 at 12:21 pm by Nathan
I'm working, nearly full time, on a Windows 10 UWP application these days. Strange, right? How many people do you know who are doing that? I bet it's not many.
Historically, my company has been a pretty strong Microsoft shop. That, plus we didn't have a true mobile strategy, left us with Windows 8/10, but on smaller devices, so it's mobile.
Oh! Mobility. Why can't we just make a web app? Let me back up.
The fuselage of the Dreamliner 787 is pure carbon fiber. It's an amazing machine, built like no other. We create, all the way from carbon threads, a huge airplane with a handful of seams instead of the traditional aluminum plates all over. Carbon fiber doesn't give way as easily to fatigue, so they say a 787 airplane we make today could fly for the next hundred years and beyond (as if the design wouldn't be technologically outpaced far sooner). Another neat trait is that carbon fiber blocks more solar ionizing radiation than aluminum airplanes, making it safer for frequent flyers and crew members. The downside is that it also blocks other radio signals very well thanks to its otherwise awesome density.
So when we send someone into one of these beautiful airplanes to do some work, there's no WiFi, most likely no 4G, and sometimes we actually need software to work while the airplane is flying (shocking, I know). Applications on our mobile devices need to be able to be offline - occasionally connected. Throughout IT history, we've found that it's easier to force a signal or make a longer wire, but we have a unique case here.
HTML5 has great offline support. Some of the things you can do are quite astounding in an offline mode, but browser limits tend to get in the way. Specifically, the limits of offline storage mean there are only so many pictures we can take and data we can collect in our mobile apps before forcing a connection. Last I checked, most browsers top out at 20 MB. Also, the Media APIs and various offline and client-side data storage mechanisms are still not stable enough or cross-platform enough to be viable enterprise solutions. Plus, internet explorer is still a thing. Ug!
So we're building a UWP app, and it's the right choice.
Googling it, however, you won't find many people doing the same. There are plenty of stackoverflow questions about WPF and XAML, but those are most all from the pre-Universal Platform days.
There are a few frameworks out there, and the Windows store has a slowly growing number of apps. It's not really that no one is doing UWP, it's really just that's it's a small community with a quieter online presence.
Maybe the Universal world is still just growing. The technology is fairly new so not everyone is on board yet. Then again, maybe the platform is far too immature to choose. We are finding out the hard way that the cutting edge tends to break skin now and then.
Recently, I found that MSTest is completely broken in Visual Studio 2017. How is anyone going to test their code if the first party framework is a non-starter? The runner up is xUnit, which works but only runs in a UWP application window. Both of those choices have the major downside of not working headlessly - they can only run from a Windows 10 PC and will not work on an integrated build server of any type because they both need a window to pop up.
We've also had debugging problems since the beginning. You can hit F12 to follow a reference out of a XAML file, but you'll never be able to follow a reference into XAML. Also you lose app insights as you get closer to the UI. Then there are the false-positive errors any time a XAML file is open, where VS is convinced that something is broken in your code while it's actually perfectly valid.
So is it a chicken or an egg? Do these problems get fixed when more folks work with UWP, or is the world waiting for some of the major holes to get fixed before jumping on the wagon? Maybe it's neither and we should all stay away. Maybe it's both and it will never start.
Most likely it will take Microsoft 5-10 years to get it right, but by then, the next big platform will come out, full of bugs.