Re: RFC: Modified FAT filesystem driver?

Albert Cahalan (albert@ccs.neu.edu)
Mon, 2 Dec 1996 10:07:06 -0500 (EST)


From: rnichols@interaccess.com (Robert Nichols)
> Marc MERLIN <merlin@magic.metawire.com> wrote:
>> In article <199611230035.TAA03352@denali.ccs.neu.edu>,
>> Albert Cahalan <albert@ccs.neu.edu> wrote:
>>
>>> I think you can always find a long filename in win*/desktop,
>>> or whatever they call it.
> [SNIP]
>> My point is just that this kind of autodection would have _many_
>> reasons to fail. For many people, a mount option would be the only
>> reliable option.
>
> Exactly. There is no requirement that a VFAT file system actually
> contain any files with long names. What would happen if you were to
> 'rm' the last long-named file from the file system and unmount it? When
> you mounted it again would you want it to come back as FAT instead of
> VFAT (and now you can't re-create the file that you just deleted)?

It does not matter if there are many reasons it can fail. It is a
heuristic, and heuristics do fail sometimes. You can always override
the heuristic by mounting as vfat, msdos, umsdos, dmsdos, or uvfat.

If there was a "fat" filesystem that would autodetect, Linux
distributions would not need to worry the matter. When the
distribution assumes a plain msdos filesystem, Windows 95 long
filenames get messed up. When the distribution assumes a vfat
filesystem, DOS scandisk will complain.

Note that the Windows 95 installer does not modify /etc/fstab!
If Linux would autodetect the filesystem subtype, then there
would be no need to modify /etc/fstab. Oh, let's try to not
torture newbies too much. Expecting a newbie to modify /etc/fstab
after a Windows upgrade is like expecting them to hack the registry.