Re: (reiserfs) Re: patch: reiserfs for 2.3.49

From: Alexander Viro (viro@math.psu.edu)
Date: Wed Mar 15 2000 - 21:33:46 EST


On Wed, 15 Mar 2000, Hans Reiser wrote:

> "Albert D. Cahalan" wrote:
> >
> > Plan 9 namespaces are like this:
> >
> > The clone() system call gets an extra flag. This flag is used
> > to control namespace sharing. Current Linux always shares,
> > because it can not support the unshared case. BSD and SysV also
> > have this limitation.
> >
> > When you don't share, you get separate mount points. The two
> > processes can mount and unmount filesystems without seeing
> > each other's results.
> >
> > For example, the login-related software (getty, login, sshd...)
> > could perform a namespace-splitting clone() at some point,
> > then mount the user's home directory right on /home. Every user
> > would see their own files /home/.profile and /home/public_html,
> > because every user would have their own namespace.
> >
> > On a Plan 9 system, users typically union-mount their own
> > executables (perhaps stored in /home/rc) on /bin. The shell
> > only looks for executables in /bin and does not need $PATH.
> > Without Plan 9 namespaces, this would be a huge security hole.
>
> I assume you are implicitly asking Viro's question. The plan 9 folks did not
> invent the term namespace, they used an already existing term from papers
> previous to them. I'd have to go look at my old thesis to find them, my advisor
> made me read them because these papers propose namespace unity and use the term
> namespace, not because they propose a particularly interesting syntax for doing
> it. I suppose I ought to be citing them on my web page. As I define

        _Please_, do it - that would solve the whole problem. Nobody argues
against using the term in the meaning you are using it. The _only_ problem
is that your text refers to Plan 9 definition and uses totally different
one. If you replace the reference to Pike et.al. with one that would not
give conflicting meaning of the term - all problems are solved.

        For anyone who knows about Plan 9 the combination "Plan 9" and
"namespaces" immediately implies "namespace as a binding of unified tree
from individual filesystems (as in UNIX), but unlike UNIX different
processes can have different namespaces". So the first reaction on pointer
to Plan 9 papers in your text is "HUH? How the heck does he do it inside
the filesystem driver", the second "Hey, he describes something completely
different, where does he see any connection with Plan 9 meaning?" and then
come, erm, unflattering ones...

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/



This archive was generated by hypermail 2b29 : Thu Mar 23 2000 - 21:00:19 EST