> > I haven't tried NetBSD, yet.  Perhaps it has a similar filesystem; however,
> > I have not heard of such.  If not, I think this would be a really nice
> > addition.
> 
> I don't know if you can make your root and user partitions msdos.  Is
> this what you were referring to, Chris?  I certainly can't imagine
> wanting to run this way...  ;-)

the only problem with it is representing files with long names, and
'odd' files, like symlinks and device nodes.

the Dos 6 (is 6 the right number?) extensions which may eventually
happen to our msdosfs would solve the former problem, leaving the
latter two.  the device node problem can be solved, with a bit of
perverse imagination.  I don't think the actual 'mountroot' call for
msdos is written, but then, i don't care that much...

frankly, i think it's a stupid idea, at best, to hack the native file
system format to do something like this...

> > For what (little) it is worth, I would think UHFS would be simpler than
> > UMSDOS, given that HFS already supports long names and soft links.
> 
> The MS-DOS "file system" is well understood and documented.  Do you have
> any idea how many "reserved for finder use" flags/fields there are in
> the HFS definition?

no, but i'm sure i could go home and look it up...  8-)

> Or what, exactly, you'd do with split data/rsrc
> forks

this one is easy: have a mount option that pick between applesingle
and appledouble formats...  they've gotta be documented _somewhere_
(probably A/UX header files/man pages, or a tech note...)

actually, i should say "the answer is easy;" I'm not sure how bad
the implementat ion would be.  8-)

> and the fact that HFS and UFS have different reserved characters
> for their filenames?

this isn't that hard...  UFS reserves NUL and '/'.  HFS reserves ':'
(that it?  last i saw, it was), and i can't think of how to get a NUL
into a HFS file name sanely.  so just map : to / and / to :...


as you've said: "NOT NOW!!!"  8-)


chris