Yes, :crlf processes unix style files correctly, but for the wrong reason. (It does not translate Unix record separators, it fails to recognize them at all.

I still don't quite follow - AFAIK, on *NIX, Perl's "logical" \n is always LF, so I don't really see any problems if it's just CRLF vs. LF files and only Windows and *NIX are involved; if there are other OSes or formats involved, then one might indeed run into issues there. But of course I'd also advise to always know one's input format (that gets especially important when Unicode is involved), so knowing whether one's input is CRLF vs. LF is a good recommendation of course. (In this particular case I think the :crlf layer gives more flexibility over input formats.)