> Stephen Frost <sfrost@ns.snowman.net> said:
> > On Sat, 9 Oct 1999, Horst von Brand wrote:
> > > IF a real replacement shows up, that has significant benefits for a
> > > substantial number of users, and which doesn't bloat the kernel, I'll use
> > > it gladly.
>
> > What are your feelings on the talk that's been happening with regard
> > to thoughts of a larger major,minor set, and a file in /proc to notify user-
> > land of what devices are installed, and when devices are added/removed?
>
> Sounds quite sensible. Minimal bloat, can just forget about it if you don't
> want to use it at all.
>
> Security implications worry me: When a new device shows up, the devd starts
> some proggie to configure it (permissions, ifconfig, whatever). This will
> be a program with root privilege, or at least device creating/configuring
> capability. OK, whom is this GUI (Big mess! Hard to audit!) program going
> to contact? Current user at the console, only user on the machine (some
> vermin over the net, perhaps), wait for root to show up? Might be OK for a
> personal, non-networked machine. For a server it is out, AFAIKS. And then a
> substantial chunk of the proposed functionality is lost.
Point being, it's all in userspace, so you can do what you want with
it. As you mentioned, you can ignore it (server, unlikely you're really
going to be changing things very often anyway), or have a daemon ask the user
what he wants to do (client, cute GUI, whatnot), or have something between the
two. Perhaps a daemon running as root that creates nodes based off of a config
file (Once the nodes are created a root user can go off and change the permissions
to his liking if he needs to). Said daemon could also delete nodes based again
off of a config file.
All user-configurable, all user-space, so users can do what they want.
This will also allow different people to come up w/ their own daemons, and
different distributions can use the one they like, or create their own. Or they
don't have to have one, and can continue with things as they are.
Stephen
-
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/