On Wed, 2007-11-21 at 19:26 -0500, Alex Jacobson wrote:
Ok, I'm game to default to haskell98 in the presence of ambiguity, but in most cases the extension involves new syntax and that should be enough.
In these cases ghc does generally give an error message which mentions which extension it is that you should use. This is actually better than the case where you forget to import something when ghc doesn't helpfully tell you which module you forgot to import. As others have said, one major reason for declaring extensions is for portability.
Note: I would also argue that extensions that change the meaning of old code are prime-facie a bad idea. It means that, if you start using the new syntax in your existing code, you have manually to audit your code to make sure nothing else got broken and that is difficult even for experts to do really well. The example you provide is actually evidence that this bang-pattern extension is probably a bad one.
In general I'd agree. In the case of bang patterns it's actually very hard to construct examples where the code compiles both ways (and is actually used). Duncan