"Pablo Aguilar"
While I haven't tried it myself, for the same reason (not available on VC6 nor BCB), I do think it'd be easier to use than trying to customize tokenizer. Also, (and again, I haven't tried it) I believe Spirit 1.6.x (spirit.sourceforge.net) is available for older compilers; I know it's supposed to work for VC6, but I'm not sure about BCB.
While I can appreciate the usefulness of Spirit, I just don't see the need for it here -- especially since it means bringing in an entire library just for a single operation in my application. In the event that tokenizer is incapable of handling my desired requirements, I have come up with a reasonable work-around: I simply pre-process the string, substituting the '+' operators with some un-printable character (like '\x01', or any other character the user is highly unlikely to type), while leaving any '+' symbols within quotes alone. Then I use the tokenizer with the '\x01' as the token separator, and voila! - Dennis