I suppose that's right, but don't you also need to know how to write VFP programs?

But what are you comparing that with? A useful comparison might be:

1) A newbie creates a VFP form, adds a Remote View to its datasource, adds dataaware controls, downloads samples from UT or wherever to add navigation/search features, then selects "compile to NET" in their design interface and ends up with an assembly that uses Silverlight and the EF. Then the developer steps through it in the debugger, adding functionality including error trapping- again based on tried-and-tested best practices documented over the last decade. They recompile for WPF and decide to use the WPF version.

2) A newbie developer starts experimenting with EF, Silverlight and WPF. There aren't many samples online for Silverlight 2 because it is so new. They end up developing a Winform using typed datasets and Stored Procedures since most of the current online advice/examples support that decision. To use the EF will require a rewrite. To use silverlight will require a rewrite.

The key advantage I see is the one most often assigned to a 4GL: it encapsulates the complexity. Why not? For me it's like an automatic gearbox: I suppose I have more control with a manual box, but the cars are so much more sophisticated and smart these days and automatic boxes so capable that I'm almost certainly better off with an auto. If some people think that makes me a wimp or not prepared to learn to use a manual properly to extract all its value, I can live with that.