dhclient starts on all interfaces referred to in dhclient.conf regardless of command-line parameters

[David W. Hankins]

On Tue, Jun 14, 2005 at 12:43:09PM -0700, Ted Lemon wrote:
> Just to be clear, the way I think it should work is this:

I'll agree and add to this the other modifications I've suggested;

>       * If you specify no interfaces in the config file or on the
>         command line, you get all interfaces on which DHCP makes sense
>         (generally speaking, all broadcast interfaces).

	And they use a network I/O method that will result in RFC2131
	compliance (notoriously in regard to broadcast address), or what
	is available, whatever that means for the build environment.

>       * If you specify interfaces in the config, and no interfaces on
>         the command line, you get the interfaces you specified in the
>         config, and no others.
> 
>       * If you specify one or more interfaces on the command line, you
>         get only those interfaces, regardless of what it says in the
>         config file.

	If a network I/O method is specified in config, it is used,
	otherwise the default is again the best method available to
	acheive RFC2131 compliance.

It's plainly visible at this point that part of my reason for eliding
this to a feature release is combining effort.  This is much easier to
do if you're already planning on making sweeping changes to related
sources.

> I consider it a bug that it doesn't currently work this way.   However,
> as I say, David is the authority here, not me.

But I also can't speak for what was intended when it was written,
Ted is often the authority there, and more often than not I'm just
guessing.


Again in the interests of being clear, I agree that this is not the
desired behaviour, but I do not presently believe that this is
appropriate to address in a maintenance release.

For better or worse, this 'undesirable feature', and we do have a
few others, is it's current behaviour, and it's possible that it
may be relied upon.

People allow us to amaze them when behaviours change between feature
releases, and scripts or methodologies they devised for them break.

People are not so forgiving with maintenance releases, particularly
for features/behaviours that have been in the released software for
a nontrivial number of years.

The exception is the severity of the problem - security, crash bugs,
protocol violations...


As always, I'm open to arguments to the contrary, but the last time
dhcp-server spoke about this no one objected, to my memory.

-- 
David W. Hankins		"If you don't do it right the first time,
Software Engineer			you'll just have to do it again."
Internet Systems Consortium, Inc.		-- Jack T. Hankins