Re: Patch for unicode support on Win32 in GNUstep base

[Adam Fedor]

On Aug 25, 2004, at 2:41 AM, Richard Frith-Macdonald wrote:
> My feeling is that, if the filesystem names are going to be unicode 
> (ie contain characters which can't be represented in any one single 
> byte characterset), the encoding should be utf-8 and so 
> -fileSystemRepresentation would return a utf-8 sequence, which could 
> then be converted to 16-bit unicode if/when it is passed to a system 
> call.
>
> The only problem with that is ... what if people set their default 
> encoding to be (for instance) latin1, but they try to use filenames 
> containing characters which can't be represented in that characterset 
> and need utf-8?  I think this could be a problem on any system though, 
> not just windows.

Yes it is. From the documentation, the methods should return UTF8 on 
systems that can handle UTF8 filenames (I think almost everything can). 
Right now I think it only does that when the default encoding is some 
form of UTF8 (perhaps that's OK, I think?))

On Windows, I had thought about making sure that happens and then doing 
the conversion to 16-bit before the system call. That's OK if we are 
just using it internally, but it would be confusing for a developer on 
Windows using fileSystemRepresentation, expecting it to return a true 
16-bit unicode when it really returns utf8