why

[Walter Bright]

John Reimer wrote:
> Well, it may be good advice, but I consider it impractical since it's 
> not at all consistent with the D language itself (where case sensitivity 
> is the norm within source files).  I think it's an ugly and limiting 
> requirement (or convention).

It's a convention. But one will run into difficulty if doing things like 
naming one module "foo" and another module "Foo".

> The naming of namespaces should be parallel to any other D convention. 
> If this is a problem, then mapping D namespaces to the filesystem was 
> not such a good idea, because this makes D more dependent on OS 
> filesystem technology and schemes.  That probably is not a future proof 
> approach.

The convention will work with both case sensitive and case insensitive 
file systems. I don't foresee any other convention ever being used, but 
even if one is, I expect it can be dealt with.

> Furthermore, contrary to the statement Walter makes there, I don't know 
> of many file systems that are truly case insensitive, at least 
> pertaining to those OSes that D is available on currently: win32 is the 
> closest that comes to that.  Even windows, though, allows files and 
> directories to be named with uppercase and maintains the case state... 
> Thus, as long as you remember to be consistent with filename casing, 
> using uppercase in module names is not an issue on Windows.

That's true, but you still can't have a "Foo" and "foo" coexist.

> I consider the residual case insensitivity in Win32 a defect and a nasty 
> side effect of the DOS era.  If Windows is the only OS that makes 
> keeping Walter's convention a "good idea", I'm not convinced. :)

It's been a while since I've used VMS, but it was also case insensitive. 
So was OS/2. DOS, VMS, Windows, and OS/2 all trace back to DEC operating 
systems which were case insensitive. Linux, etc., all trace back to Unix 
which was case sensitive.