>If the C compiler does NOT do some consistent normalization, then >the actual IDENTIFIER that the linker tries to resolve will not >match and the link will fail.
You are assuming that careless programmers have a mix of source code that is not self-consistent. If the convention is to use NF-C and the programmer generates a bunch of NF-D code which he tries to link to NF-C code, then a most I would say give him a warning from the compiler, then keep going. If the link fails, well, its supposed to. >To assume that some keyboard input method does the normalization >is naive. Lots of C programs are machine-generated (more and more >in these days of UML -> WSDL -> whatever language tools). Like I said, if you want to generate source compatible with other people's source, just use the convention they do. The compiler/toolset doesnt have to care about it, and its better that it doesnt. Now, on the other hand, I could imagine a tool which takes a text file, and applies an "ANYUTF8 -> C/C++_NORM_FORM" transformation on it, as a convienience for those with poor input systems. But definitly not mandatory, because you may just want to write a wacky program which has lots of non-normal forms in in on purpose, and there is no need for the compiler to contrive a failure. -- Linux-UTF8: i18n of Linux on all levels Archive: http://mail.nl.linux.org/linux-utf8/
