With all of the discussion of HFS and UMSDOS, why not make a UMSDOS-type
filesystem on HPFS? It already has full support for long filenames. [...]
Mark Lehrer <lehrer@bellhow.com> writes:
yes! good idea.
i think the standard HPFS should work via extended attributes for
permissions, group, owner, etc. no special UHPFS would be necessary.
Yes, HPFS can do it all except hard links. But in my opinion, that's
enough to make it a bad idea to do all the rest -- since it can't be a
primary Unix filesystem, the fact that e.g. device special files
>could< be supported correctly doesn't mean a lot, and putting code in
the kernel to do that would be wasted.
Something like UHPFS could be done more easily than UMSDOS, but why?
It wouldn't be better than UMSDOS (whose virtue is portability) and it
wouldn't be better than ext2 (not without repeating the effort that's
gone into ext2).
Mark Lehrer <lehrer@bellhow.com> writes:
[...]
there was talk when 1.3.x was young about read-write HPFS being
one of the goals of this kernel version; is this still the case?
"James L. McGill" <fishbowl@pic.net> writes:
I had assumed that the reason for no HPFS write support was that
IBM or whoever would not release data, or something. I assumed
that, since there wasn't even an experimental, dangerous,
guaranteed-to-trash-your-filesystem read/write HPFS.
[I wrote the read-only HPFS, and just stumbled across this discussion.]
I can't speak for anyone else, or even quote what I've heard because I
haven't been following too closely, but by being willing to risk freaking
out HPFS386, whose addons to the data structures might have almost
unlimited ramifications, and by being willing to risk freaking out CHKDSK
after a Linux crash where a bunch of partial updates might be a new
experience for it, someone could probably write a robust enough HPFS
writer. The lack of documentation from IBM and MS is probably not fatal.
In my case, though, simple intellectual interest hasn't been enough to
carry me through the massive amount of code that needs to be written
(I don't actually need a writable HPFS for much), and possibly something
similar is true for the guy who's been working on it recently.