Since the pure perl is the fallback solution, and it does not require anything but perl (and well, if you're trying to install Text::CSV you probably have perl on your system), it's not really surprising to see CSV_PP is provided with Text::CSV (it's not installed at the same time, it's part of the module. So you can't have CSV_PP without Text::CSV being installed (or it was done on purpose for some reason)).

I don't see anything preventing Text::CSV_XS being installed without Text::CSV, so indeed, this might be an issue. But Text::CSV with its PP implementation is a fairly simple module: only to .pm files, without any compilation needed (that's actually why it exists, in case the compilation for XS fails), so it should be easy to install even "by hand" (unzip and perl Makefile.PL; make; make test; make install), or even just include without any installation in your program. Besides, its main benefit is that the implementation is abstracted away, and you get one unique interface for all your calls, so no my $csv = Text::CSV_PP->new vs my $csv = Text::CSV_XS->new, you just use Text::CSV and everything in the Text::CSV namespace will point to the right element, which was your second problem :) .