We're starting a new project and naturally we looked at leveraging the latest .NET framework features (auto properties, extension methods, lamdas, LINQ, etc.). The question of user interface came up and we had some decisions to make.
This specific project we looked at building one client web front-end (for the majority of the users) and a SmartClient for a smaller contigent that requires a little more real-time feel along with more rich features the web client might not be able to do (without a lot of heavy lifting). As we were looking to start things, the notion of WPF vs. WinForms came up. This led us down a path to look at Prism (the Composite WPF project from the Patterns and Practices group) which has a lot to offer in frameworks and being able to do more with WPF (Prism is WPF only). WPF also has some benefits to automation (there are some new products that are tying into the automation frameworks) and testing. The pretty UI is just an afterthought and not the focus since we're building what would be considered a "traditional" business application (forms and maybe grids and reports).
WPF also has the advantage that we could deliver the application using xbap which kicks ClickOnce up a notch. The Prism framework is based on modules loading into XAML regions so we were even looking at re-using XAML user controls and plugging them into different clients.
This has kicked off a bit of a controversy and discussion. Is WPF really suited to business applications? Of course the answer is (as always) it depends. If you're business application requires a rich UI and a visual paradigm, then the obvious answer is yes. It's more intuitive for a user to drag a part from one component to another (say in a room designer) than to pick items from a grid and enter values in a form.
However for the larger business applications (the meat and potatoes of the corporate world) we don't deal with "rich UIs" so where does that leave you? WinForms is not dead and according to Glenn Block is the "recommended breadth solution for LOB application development in the foreseeable future". The trade off is that if you're building SmartClients you might look toward a framework that provides a lot of OOTB features that you don't have to do. I'm not a big fan of NIH syndrome and the idea of having to rewrite loggers, event handlers, aggregators, presentation models and everything else you build into a SmartClient. Frankly, I don't get paid to build frameworks or re-invent the wheel for every application my client wants.
Prism provides a nice framework (IMHO CAB done right) that you can leverage and pull it out in piecemeal to what you need. Unfortunately, it's WPF only. While some of the principles can be carried over to WinForms I think you'll end up doing more work to say try to get DataBinding working the same way (if at all). There are other aspects to WPF (ignoring the pretty UI) that don't carry over to WinForms (routed events, event aggregation, etc.) all of which you have to build yourself (or use CAB which is the overbloated bazooka edition of WinForm programming).
The word from my twitter folks is somewhat slighted toward WPF:
WPF databinding isn't as evil as its predecessors and some people wouldn't build a WPF app without it
Control templating and eventing in WPF is useful even for the basic forms and datagrids
Testable automations ARE good reasons to use it
Other questions that come up:
Do you really need/want a separate designer to build your UI and keep it completely separate from your application code? In WPF this is easier. For WinForms you need to outfit them with Visual Studio and frankly, I've never used a designer on a WinForms project (I have on WPF ones, but not business applications)
What is the adoption from the user perspective and can you get them to think "outside of the form" and be creative with solutions. In WPF you're not as bound to traditonal approaches to visual problems and building the same solutions is somewhat easier (trust me, manipulating pixels in WinForms is a bugger to put a control on top of another one, in WPF it just works)
Can you bake the learning curve of WPF into a business application budget or should it be something that developers just know? (i.e. only start a WPF app when you have developers that know it inside and out, which few do these days)
I’m on the fence on WPF. From an automation and programming model, I think it’s fathoms above WinForms. I don’t mind the crafting of a good looking UI as I’m used to it in Blend and can whip off a UI just as fast as I can with WinForms. However I’m not sure its worth the extra effort (is there extra effort) in building business applications with it, but I’m torn with xbap where we can deliver a rich user experience without having to install a client.
For WinForms you generally have to either harvest what you have from existing applications or RYO when it comes to the application framework. Re-writing presenter base classes, validation strategies, various implementations of patterns like repository and specification, and building a service layer over things like NHibernate can be fairly straight forward but still might take a few weeks, weeks of cost that the customer doesn't want to pay for to see nothing out the other end.
There is the build as you go model, which works for any technology however you do want to keep consistency when you're building 5+ large applications a year and there's turnaround in the IT department. Re-inventing the wheel for every application isn't an option for anyone.
Feel free to chime in with your thoughts, experiences, and ideas. Thanks!