tthorn@daimi.aau.dk: "No VFS means no FFS?" (Jan 13,  3:53):
quoted text
>    From: Linus Benedict Torvalds <torvalds@cc.helsinki.fi>
> 
>    Ok, as I didn't implement the VFS-patches (they seemed a bit raw), I
> 
> I haven't seen the VFS-patches, but as I understand, they are
> essential for the Berkeley Fast File System.

The VFS that is available as alpha (tsx-11, ALPHA) has the problem that
it wants you to rewrite almost everything when you add a filesystem (ie,
have a minix_open, and a ffs_open etc), which I find a bit boring, and
also makes for somewhat ugly code (several versions of essentially the
same set of functions). I decided that I didn't want to lock into that
yet: it might still be the way linux goes, but I don't want to have an
official release that binds me to something I think could maybe be
implemented cleaner.

VFS will come, but I'd rather see it (I'll implement it too, perhaps) as
a collection of just the low-level routines (read_inode(),
write_inode(), read_dir_entry(), write_dir_entry(), _bmap() etc), where
the general routines like "link()" etc would be the same for all
systems.

		Linus