How large and economically-successful a startup are you interested in?
I'm bootstrapping a native-client-software-based business. My business isn't successful yet, but we'll see how things play out in the next couple of years. I'm hoping that though web-based apps are all the rage, there still exists a large-enough market for desktop software. My economic goals are modest (though a wildly-successful outcome wouldn't be unwelcome).
For several years I've kept up with many, very-helpful startup-related blogs and sites. The vast majority of these recommend launching web-app-based startups. Even one startup owner who had hybrid approach phased out his desktop version (http://www.kalzumeus.com/2009/09/05/desktop-aps-versus-web-a...) So why am I developing desktop apps? The main reason is that I care passionately about user interfaces and the effect they have on my experience as a user. There doesn't seem to be a very consistent UI for the Web as a platform, not as consistent a UI as exists for Windows, Mac OS or GNU/Linux distributions. Web apps have their advantages, but UI-consistency isn't their strong suit (nor do I foresee this changing very soon).
A second reason for my focus on desktop apps is that I want to scratch my own itch and make programs that I want to use myself. I want to write high-quality software that 1) aids me with tasks related to various personal interests and 2) is desktop software, if possible. I want to exclusively produce "dogfood" (though not necessarily programmer tools). Having myself as a target user fuels my motivation to get up and work hard on making a streamlined, powerful, easy-to-use product, since I'll be using it. Yes, I'll still perform usability testing and try to understand my other target users. However, I believe it can be helpful to be a target user myself. I can't imagine how hard it must be to develop software for an unfamiliar field or making a program whose only value to me is as a source of income. (By the way, having oneself as a target user is generally considered by startup bloggers to be a mistake.)
A third reason for my desktop-app obsession lies in the opportunities I see to leverage the advantages of native client applications. Various articles lists the advantages of web apps. However, desktop apps have their own particular advantages, though many desktop applications fail to leverage them. I plan to keep identifying these advantages and capitalizing on them. At the same time, there's a lot for client apps to learn from web apps (and I hope to internalize those lessons as well).
As you can see, I'm a bit of a contrarian and somewhat stubborn about this particular issue. It may turn out to be my undoing. (I hope otherwise, of course.) Check up on me in a year or two... In the meantime, be careful about any advice you may draw from what I just wrote.