Re: Basic Question.
James Carlson <carlsonj <at> workingcode.com>
2005-12-07 13:10:56 GMT
Bill Unruh writes:
> One of the real problems with linux appears to be the difficulty of
> associating a unit with a physical device. Thus with ethernet, it depends
> on which hardware device the system just happens to look at first which
> gets called eth0.
OK, but it's also the case that pppd is *more* than just Linux. In
fact, it predates Linux by quite a bit. So, I'd argue that the
solution needs to work for other systems (BSD, Solaris) as well as
Linux.
> The aliases established in /etc/modprobe.conf seem to be
> just ignored. Similarly with ppp, and everything else. This really
> needs a unified way of dealing with it. the unit option in pppd helps, but
> if for some reason that unit is unavailable, ppp simply grabs whatever
> anyway.
>
> To have to set up a script which determines which units are which and
> rewrites all references to the unit number is not only a kludge but is also
> extremely prone to bugs. That is not the way to go.
I agree that's kludgy. One of the bad parts of it is that the
interface is established and enabled *before* the filters get applied,
and that's almost certainly not what anyone concerned with security
wants.
But even if we could somehow nail down the meaning of "ppp0" and
"ppp1" and so on, I think we'd be no better off than we are today for
pppd. The problem I see is that a system with a number of links and a
large number of peers (not just this simple configuration with two
(Continue reading)