Re: Case Insensitive File Systems -- Torvalds Hates Them

On Tue, 29 Apr 2025 19:04:49 +0000
Farley Flud <ff@linux.rocks> wrote:

On Tue, 29 Apr 2025 20:11:19 +0200, Andreas Dehmel wrote:
 

On Mon, 28 Apr 2025 18:56:18 +0000
Farley Flud <ff@linux.rocks> wrote:
 

On Mon, 28 Apr 2025 11:12:42 -0700, John Ames wrote:
 

 
Just so, it seems to me. Of course it's many years too late for
*nix to course-correct on this, but it was a stupid design
decision in 1970 and it remains stupid now. Well, such is the
nature of things in this vale of sin and tears...
  

 
Case insensitivity was only idiotic at the beginning, but now, in
the age of Unicode, it is supremely idiotic.
 
Consider the German "sharp s," which I cannot enter as UTF-8 here.
 
But the lower case sharp s maps into TWO DIFFERENT upper case
chars: <can't enter> and "SS," e.g. STRASSE or <can't enter>. 

 
That merely illustrates the point that whoever decided to model it
like this in Unicode was truly a numbskull. For two reasons:
 
1) just because the result _looks_ like SS doesn't mean it has to be
two characters. A Unicode character can look like anything, even a
full word (and beyond). The only reason to use two characters would
be hyphenation, which in this case is explicitly forbidden. Someone
didn't understand the difference between syntax and semantics.
 
2) this transformation is not trivially inversible. No, you can't
just translate every SS back to ß, you'd pretty much need an AI to
invert this. Whenever you're introducing a transformation that's
trivial in one direction and extremely hard in the other, and
you're not working in cryptography, you're doing something
extremely, horribly wrong.
 
 

There are special rules on case folding for thousands of Unicode
chars and the "sharp s" example is one of the simplest. 

 
I seriously doubt that, especially since many (most?) languages
don't even know what "case" is supposed to be in the first place
(such as Japanese, I'm pretty sure it's the same in Chinese and
most other asian languages, which incidentally take up the most
code points). And even if it were true, that'd mean we'd need a
couple of thousand additional code points for these special cases,
out of several million -- who cares, the
gender-neutral-smileys-crowd?

 
Thanks for your input.
 
I am not a native German speaker and I can only rely on web sites
to inform me of these issues.
 
But I don't quite get the thesis of your post.  Are you for or
against case insensitive filesystems?

The point was mostly that it annoys me when people make really bad
design decisions like the non-inversible ß -> SS transformation
(which keeps getting repeated whenever case-issues come up) and then
complain that for some inexplicable reason there are nasty repercussions
down the line. This stuff _could_ be really simple; it's almost as if
there's a lesson in there somewhere.

As far as filesystems go, I personally don't have any strong feelings
about this either way, mostly because when I got into computers you
were lucky to have an ASCII equivalent. But I can tell from long
experience as a developer of an application for engineering types that
even those guys usually can't be bothered with case-sensitive
filenames, never mind the average Joe. Nerds might see this differently,
but the vast majority of computer users are not nerds, and in my
experience things like case-sensitive filenames are a constant source
of irritation and confusion for everybody else.



Andreas Dehmel