Re: umsdos/uvfat

James Mastros (root@jennifer-unix.dyn.ml.org)
Mon, 2 Feb 1998 01:32:20 -0500 (EST)


On Sun, 1 Feb 1998, Toby Reed wrote:
> > > 1) not all world has converted yet to Win95, some still has only old DOS
> > > partitions around. Which work without ugly vfat thingies.
> > So? We convert them to vfat. All but the most low-level (defragmenters,
> > fsckers) will work fine.
>
> not everybody likes vfat. I don't even know if it works on floppies or
> not.....
It does.

> besides, I like to use low-level stuff on dos partitions. I have never
> installed Windows 95 and I don't intend to...it's long file names just
> don't attract me.
Nobody is asking you to.

> there are not lfn-aware replacements for some of the tools that I use in
> DOS, a few dating back to 1988, etc.... unmaintained.
There may well be... mind naming a few of them, and what they do? I'm
betting that most of them have Linux equivlents.

> > > 3) It's been here for a while. There are distributions that run off the
> > > msdos disks, that run under umsdos. Didn't seen any that does under (u)vfat.
> > Yeha, so? They can stay with 2.0, or they can upgrade to 2.2 and uvfat (not
> > all that hard, really. If nothing else, have a bootstrap with all 8.3 names
> > that extracts the rest with info-zip for DOS (does lfns), or under Linux).
>
> So vfat is really really not a good solution for everyone. UMSDOS has been
> here since the beginning and before we commit ourselves to using a
> Microsoft standard (vfat) compared to a Digital Research standard (fat),
Interesting... vfat is fat, just using some formaly illegal attribute
encodings to add more times (ctime and atime), and long file names.

> both of which are horrible, the former even worse,
Odd, as vfat is a superset of fat.

> lets have a look and
> realize that many people are dependant on umsdos and converting to vfat
> only adds another layer of annoyance.
And removes a layer of annoyance. Files show the same names and times in
win95 (if you have it) and Linux. There is less dependance on the
---LINUX.--- file.

> I can't stand dealing with vfat since whenever I trade files with anyone I
> end up with Fejakoea.~1
That depends on how you translate filenames. I should think that we would
want to do the NameNumericTale=0 equivelent... that is, keep the first 8 and
the last 3 characters of the file as the shortname.

> The extentions don't match and the file doesn't show up. UMSDOS takes care
> of this beautifully, moving from vfat to fat to floppy to unix, no
> filenames are lost.
As will uvfat.

> vfat on the other hand, mangles everything into file.~# and defeats the
> whole purpose (preserving the long names AND directory data, gid,uid,etc).
No, it does nothing of the sort. It mangles filenames by talking the first
8 characters as the filename part (ie the part before the dot), and the
first three characters after the last dot as the extention. That is to say,
the filename is mangled as little as possible. A ~# replaces the last 2
characters of the filename part only if it is neccessary to produce a unique
shortname.

> If I had my way everyone would be ext2, but uvfat is not imho a viable
> alternative for everyone to umsdos.
Hmm... I really don't think that that is true.

> Converting would mean these things:
> - I probably would have to dirty my system by installing 95 to do the
> actual convertion, and having 8 or 9 dos drives it would take a while and
> some amount of effort.
Umm, no. Vfat IS fat, no conversion is neccessary.

> - I would have to replace my old trusty assembly dos utilities with icky
> bloated ones that support lfns
No. Do it under Linux, with trusty new C dos utilities.

> - I would lose a large amount of data in the form of filenames and
> information when moving from floppy to zip disk to hard drive to fat drive
> to unix, etc
No. They would work exactly as now. Indeed, umsdos can be thought of as a
special case of uvfat.

> - without some sort of special convertion utility all my existing umsdos
> filenames and permissions would be lost
True. So run the utlity. It isn't at all complex. Baselene:
1) Mount the partition as vfat.
For each directory:
2) Read the ---LINUX.--- file.
3) Rename the files to the names specified in the ---LINUX.--- file.
4) Set the atime/ctime as specified in the ---LINUX.--- file.
5) Rewrite the ---LINUX.--- file

> - I would have to deal with the non-case-sensitivity of dos and long file
> names
No, you don't. VFAT stores names with case: it is just that windows 95 will
match without regard to case. The vfat fs follows suit (now), but there is
no pressing technical need for it to (indeed, it is easyer for it not to, as
that is what the default d_ops do.)

> blah blah blah
Foo! Bar! Baz! Qux! bOb! There is no God but Aries, and J.R. Bob Dobbs is
his prophet!

> my point: umsdos isn't dead yet..........just fix it and don't worry about
> it
Hmm... How about this:
We create a filesystem, call it unat (UNix fAT).

It would work like this: file times and names are storable in both the vfat
and umsdos ways, with a config switch to pick (defaulting to just vfat).

File types and names are readable in both the vfat and umsdos ways, with a
config switch to define which is prefered.

Permissions are always stored in the umsdos way, as vfat dosn't have room
for them.

I would suguest that, to start with, setting read=vfat,umsdos,
write=vfat,clear. (Clear=remove redundant ---LINUX.--- name/times.) Then any
action on a file would result in it being updated. You can run a simple
shell-script to update everything to the new scheme, but if you don't,
everything will work. read=write=umsdos does as umsdos does now. This
would be a lot more work then doing a straight fix of the existing scheme,
but is much more flexable, and yeilds only one bad-hack fs.

> -TR
-=- James Mastros

-- 
   "I'd feel worse if it was the first time.  I'd feel better if it was
   the last."  
   	-=- "(from some Niven book, doubtless not original there)" 
	    (qtd. by Chris Smith)