Andrew Sackville-West wrote:
On Thu, Oct 25, 2007 at 10:35:10PM -0400, Douglas A. Tutty wrote:
On Thu, Oct 25, 2007 at 01:23:38PM -0700, Andrew Sackville-West wrote:
On Thu, Oct 25, 2007 at 08:48:39PM +0100, Joe wrote:
Douglas A. Tutty wrote:
Now, if you actually had a piece of hardware that _was_ fully supported
by the linux kernel without this mess, then you would get a functioning
eth0 which would then work just fine with the standard Debian networking
tools. In short, your problem isn't with the networking tools, its with a
non-functional driver.
Looks like my last post didn't make it here. I had a solid, dependable
working MB NIC until a couple of weeks ago, when Sid suddenly started
renaming it to eth1 during boot (without explanation) and then saying eth0
didn't exist. I didn't notice (not exactly the kind of thing you expect),
disabled it and installed a PCI card, then when that didn't work, had to
alter my interfaces file to eth1. I think the MB NIC is probably not
faulty, but I'm short of time at the moment and it isn't urgent.
Not exactly standard Debian behaviour, or at least it wasn't once.
well. I'm silly for jumping into this, but the important word above is
'Sid'.
I suspect that's 'nuff said. :)
There's probably some way to get udev to be consistant with device
names. If you know the module that gets installed, I wonder if putting
it in /etc/modules would cause it to name consistantly.
Also, I vaguely remember that pre-udev, pre-devfs, there was a way to
identify the unit-number based on MAC address when the module was
loaded. I think it was aliases in /etc/modules.conf. That file doesn't
exist on my box. So I guess you'll have to learn about udev to get
persistant naming.
I think there is pretty good persistent netowrk interface naming at
this point. I was just pointing out that the OP above was complaining
about it not working in sid... which is pretty standard expectation
for Debian behavior.
Regardless, though, its pretty straightforward to get persistent
naming provided you can get some bit of unique info about the device
out of the kernel. Once that's done, you *can* make a symlink with a
good name that will always point to the desired device, regardless of
what that device might be named...
not at my usual machine to provide a link to the udev rules tutorial,
but its pretty easy to find.
Yes, I know what to expect with Sid, which I've been running since
before Sarge was released, and this isn't it. Sid is for incorporating
new software variants into a future Stable, and sorting out any
integration issues, not for troubleshooting broken software. It's
supposed to work *before* it arrives in Sid.
I had one (1) Ethernet adaptor, which used to be called eth0. I can
understand potential confusion if there was more than one, and I've seen
that happen. What conceivable reason is there for designating it eth1,
the second Ethernet interface, when the machine contains only one?
Workarounds there might be, but why do I need them?
The thread is about the wisdom of setting up networking in The Debian
Way. The point I was making is that The Debian Way today clearly isn't
The Debian Way of a month ago. It used to involve editing a text file,
and at worst tweaking the modules a bit, now it involves learning the
operation of an entirely automatic system that the user isn't even
supposed to see, and how to override it when it screws up. As far as I'm
concerned, that's The Windows Way, and it doesn't belong in Linux.
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
