Not to take away your motivation, but parsing CSV like this is a dead end. There are way to many edge cases to make this work reliable (if at all). There are two de-facto CSV parsers available already that support all the options users will ask for: Text::CSV_XS and Text::CSV (which uses Text::CSV_XS when installed).

If your CSV restricts itself to the simplest CSV possible - comma separated data, no embedded newlines, no mixed line endings, no unicode only supporting the default quotation and escapes (both being ") - then you can get away with the dead-simple approach, but Text::CSV::Easy_XS and Text::CSV::Easy_PP are already available for that.