(In reply to Christian Boltz from comment #5) > (In reply to Noel Power from comment #3) > > We discussed it here and decided that maybe a non printable delimiter is > > probably a better choice (and avoids mostly any chance that it could be part > > of a path name scanned by sed) so we chose the first single byte ascii char > > that worked e.g. 001 > > That's an interesting delimiter ;-) which I never used, and that might make > the script a bit more interesting to read. most editors (at least gedit, emacs & vi) show it as a special char > (And I won't even think about copy&paste ;-) that might be a problem (but not really necessary and in editor cut/paste should work) :-) > > To make things a bit easier - the delimiter should be avoided (or escaped) > in the sed command (the 's/foo/bar/'), but it shouldn't be a problem if it > appears in the input data (in this case the path name). This also means > using ';' as delimiter (as my patch does) should work without problems. > > If you want a quick confirmation of the above: > echo 'test;delimiter' | sed 's;t;b;' right, I wonder why the special character was chosen as a delimiter in the first place, we thought it was intentionally chosen to ensure it was very unlikely to be in the input data. We already have a submission with the non-printable char (as this was the reviewed change and we wanted to get this fix out asap) I doubt anyone here has any religious objections to changing the delimiter :-) especially if it appearing in the input data doesn't have downside so we will try to update this again later