On Wed, Feb 21, 2001 at 10:21:57AM +0000, Nick Ing-Simmons wrote: > So the transparency is achieved by having two representations (as ever) > > byte - cultural info from native EBCDIC > utf8 - transform and then hold as UTF-8 Unicode - cultural info from > Unicode db. This would make ord() non-transparent wrt the translation. My proposal is byte - cultural info from native locale utf8 - transform and then hold as UTF-8 locale-permuted Unicode - cultural info in 0..255 range from the native locale, cultural info in 256..7FFFFFFF range as in Unicode db, cultural info in 80000000..FFFFFFFF range as in Unicode db shifted by 80000000. [the third range is used to use chars from latin1 which are not in the locale's repertoire - IIUYC, no need for EBCDIC]. Ilya