Has anyone done much A/B testing with a rich client (i.e. WinForms or WPF) application?
I am interested in deployment strategies that make sense for this scenario.
It seems like ClickOnce deployment would be a good fit, but there's so much baggage with getting ClickOnce to work properly for a non-trivial application.
If using ClickOnce, how do you push changes for 50% of the users? Is this possible, or do you just need to essentially deploy 2 separate application?
If using Windows Installer, does it make sense to push a new installer for each test scenario? This seems onerous to the end user who has to update the application each time? Is there a major downside to installing just a stub on the client PC and dynamically downloading updates when available?