Collation implementation

[Shiv Shankar Ramakrishnan]

|implementation has been to use glibc's data files and algorithms
|(especially WRT string collation, for which I cannot simply use
|nl_langinfo), but that is not a good choice because the data file format
|is subject to change, and the algorithms themselves are in flux.

I would suggest that you have a serious look at ICU -
http://oss.software.ibm.com/icu/

Its open source, heavily used and stable with a well documented data
files format (binary as well as text mode). And I have used it a bit.
At this stage I can say that probably its the most usable I18N library
available.

Mind you that libstdc++ locales etc hasn't been finished and even then
there would be issues about what backend system locale facility to use
etc. Maybe there could be a possibility of using ICU also for this.

Thanks,
Shiv