On Wed, Apr 18, 2012 at 12:38 PM, Doran, Michael D <[log in to unmask]> wrote: > > ISO 2709 doesn't care how many bytes your characters are. The directory > > and offsets and other things count bytes, not characters. > > That was exactly my point. (Which I am stating since you quoted me and I > couldn't tell if you were refuting my point, or using it to support your > conclusion.) ;-) > Z39.2 counts octets, but say they're counting characters. If you find a record that appears to use characters instead of bytes, ignore it; it's legacy R'LMARC, which has been declared officially dead- it's Z39.2/ISO2709 that are eternally lying. UNIMARC *can* allow UCS-2 encodings in data fields, but it does not seem possible for this to imply that lengths are in characters on any charitable reading. This is because the information that UCS-2 will be used is located at a non-zero offset within a fixed field. If offsets were in character units, rather than bytes, it would not be possible to locate this value within the field. Simon