Jason Thorpe wrote:
> 
> Actually, I think ACLs could be handled in much the same way, using
> a thin vfs layer, and a db that mapped inode # to ACL :-)

Ah... then our schemes for ACLs are not at all as similar as our HSM
schemes :-)

My scheme expands the gid concept in a backwards compatible manner. It
also works with non-file gid permissions checks (SysV IPC, or whatever).

Basicly, allow users to create groups, and have these groups defined by
ACLs rather than just a simple list of users.

I'l elaborate later...