Re: [RFC] automount based devfs replacement

From: Theodore Y. Ts'o (tytso@MIT.EDU)
Date: Wed Apr 19 2000 - 17:35:54 EST


   Path: there.is.no.cabal
   From: orc@pell.portland.or.us (david parsons)

>Dynamic filesystems have their place. However, /dev isn't one of them.

        This argument has been done to death, so I'll do my executive summary
        for why it is:

        /dev is a database of the devices that are attached to the system.
        The system should publish, in a usable without black magic form,
        this database.

Well, here's the executive summary on the other side. /dev is *not* the
database of the devices that are attached to the system. That's the
fundamental devfs fallacy. It is a place where programs can "connect"
to devices which are connected to the system. In this (more traditional
Unix) view, device files are in the same class as Unix-domain sockets,
and named pipes. They are special files; nothing more, nothing less.

When you create /home/ftp/dev/null, etc. you are not creating a limited
"view" of some database; you are created a directory of special files
which are required for user programs to function correctly in a chroot
jail. Again, nothing more, nothing less.

This idea that /dev should be a "database" is not a traditional Unix
view, and given that files in the /dev directory do have user state,
such as permissions, user and group ownerships, etc. that need to
survive device insertion and deletion, it drives this point home quite
forcefully.

                                                - Ted

-
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 : Sun Apr 23 2000 - 21:00:16 EST