On Tue, 3 Apr 2001, Rick Kelly wrote:

> Konrad Schroder said:
> 
> >I think that's the point: *you* don't, but other people do.  Phil Nelson
> >does, according to his post.  I do.
> 
> The point that I haven't seen answered is why not have two file
> systems. One being bog standard FFS and the other being FFS with
> the addition of ACLs?

Because there's no "bog standard" FFS.

Our FFS already supports multiple formats. Check out man fsck_ffs where
it's talking about raising the level of a file system. Yes, by default,
newfs will make a level 2 fs, but that doesn't mean that's all our code
will support!

So we already have multiple file formats in our ffs/ufs code. When I was
at NASA, we worked with a "large inode" ffs (made by Jason), an ffs which
had inodes twice as big as normal. We really didn't have to change that
much code (it helped that the front of each inode was the same).

Also, the existing code can deal with a cross-endian FFS, when compiled
with the right kernel options. So that's a case where all of the read and
written data need to be modified before being used internally.

So since our ffs code already has the ability to deal with multiple file
system formats, I see no reason why we should not add ACL support into our
ffs code. Making it a kernel option is fine. And an ACL-FFS will just be
a slight modification to the on-disk format; either a different "level" or
a different magic number. It might also be that newfs doesn't make them by
default for a while.

Also, chances are most of the code will go into the ufs area, which will
permit lfs to use it too.

Take care,

Bill