I've built 5.8.0 using cygwin. Built and tested, out of the box. There was a complaint in the test about some *dbm problem that I didn't track down.

What's wrong with using cygwin? AFAICT, it's the best way to get a large set of CPAN modules working straight away. Most of the XS stuff has to be massaged for the VC builds and I would imagine this is the case with the Borland compiler as well. If you had a body of Perl that you needed to deploy on Unix and you wanted to do some development on W2K, I'd say that cygwin would be the way to go.

Now, if you require a lot of Win32 modules that were developed under VC, I can see that, but, generally, I have few problems with cygwin.