Thor Lancelot Simon wrote:

> However, one should note that this ioctl appears in no standard and that
> we provide a much, much less crappy way for callers to get interface
> addresses.  It is probably much less effort to fix each broken caller as
> we notice it than to even bother to keep arguing about this.

The one practical annoyment for me out of all this is that we broke
binary compatibility in a reasonably major way.  If you use dhclient to
obtain an IP address you now can't boot a kernel built after a certain
date and have that work with an older userland.  For example, if someone
boots a NetBSD 5.0 kernel (when that appears) with a NetBSD 4.0 userland
- you lose.

Couldn't we have just versioned the old SIOCGIFCONF that kept the
previous behaviour under COMPAT_40 (or some other option name) and
assigned a new ioctl command number for the fixed SIOCGIFCONF?

Yes, we do break binary compatibility every now and then, especially
with the kmem grovellers.  But this is the first time I can recall that
just putting a newer kernel on a box resulted in something I could no
longer boot because its network never got configured.  This is something
that NetBSD has in the past tried not to break - I think it sets a bad
precedent to start doing it now.

Simon.