Re: Community Input for the Maintenance and Revision of the Ada Programming Language

["Randy Brukardt" <randy@rrsoftware.com>]

"Dmitry A. Kazakov" <mailbox@dmitry-kazakov.de> wrote in message 
news:oml2tl$11i5$1@gioia.aioe.org...
>>>> On Friday, August 11, 2017 at 10:42:12 AM UTC-4, Justin Sq wrote:
>>>
>>>>> On the otherhand Ada.Directories and Ada.Text_IO are woefully 
>>>>> outdated.
>>>>> For example, if you have a computer and you speak a language like
>>>>> Japanese, French, German, Spanish, etc. think of how frustrating it is
>>>>> when you simply can't open a file with anything other than ASCII in 
>>>>> the
>>>>> path. A program that deals with relative directories might work fine 
>>>>> in
>>>>> one directory but get moved and suddenly fail hopelessly. Poor 
>>>>> end-user :(
>>>>
>>>> My mistake String supports more than just ASCII (Duh), but Asian and
>>>> other esoteric languages obvious not.
>
> I believe GNAT implementation under Linux shamelessly ignores this and 
> passes UTF-8 encoded names through.

Their claim is that it happens automatically; apparently no one (either GNAT 
or Linux) ever checks the characters of a file name for legality. So UTF-8 
characters "work", since any arbitrary code can be in the file system.

OTOH, most other systems have checks, and this doesn't work on the Mac, on 
Windows, etc.

The problem is that AdaCore's desire to keep this misfeature killed the 
first attempt to allow UTF-8 in file names generally. There needs to be some 
way to tell UTF-8 strings from Latin-1 strings (as characters over 128 are 
encoded differently); I believe there was a suggestion to use a BOM to do 
this. That would have broken the "free" use in Linux, and it was killed. So 
you guys got nothing.

The "obvious" fix is to add Wide_Directories and Wide_Wide_Directories, and 
then add Wide_ and Wide_Wide_ versions of all of the file name routines in 
the file packages. (This is the "Tom Petty" solution to me, since it would 
result is adding Wide_Wide_Open.) This Wide_Wide_ stuff is absolutely 
disgusting, and it doesn't provide a proper solution to anything UTF-8. Thus 
I'd like to find a better solution for generalized strings. I think this is 
where this discussion started ...

Round and round the merry go around we go, never to stop. :-)

                              Randy.