A partial rebuttal

If the network goes down in most enterprise environments, your desktop is a paperweight and
there's not much that NM, or anything else, can do about it.  DNS lookups will fail.  NFS
(where your home directory, your cross-compiler environment, many basic shared utilities to do
your job, your IDE or CAD or whatever software, etc. is stored) won't work.  You can't access
your email.  You can't access your calendar.  You can't work on documents which are all
provided from a central server.  You can't check bugs, which are all stored on a central
server.  Heck, sometimes you can't even get your phone working since it's using VOIP.

Exactly what magic do you propose NM will perform, that will allow a system in a heavily
networked environment to be more productive without a network than in the past?  I don't get
it.  Even without NM, back "in the day", if the network went down I could still use my local
files etc. just as before.

I definitely agree that features like Upstart are great.  The problem is, we do not HAVE
Upstart, not really.  Even after what, 2 or 3 Ubuntu releases now where it's been available,
hardly anything uses it, and certainly no network services use it.

DBUS may have been designed to be cross-platform but the fact is that it is not DEPLOYED
cross-platform.  That means for all practical purposes any change to any network service (nis,
ldap, autofs, apache, mail clients, etc. etc.) that is made for this is a special case JUST
for Linux.

As for allowing local accounts to log in, that's already the way it works: nsswitch.conf
specifies that local files are searched before NIS etc.  This is a solved problem that works
fine: we don't need NM for this.

I also don't understand your point about DHCP and leases.  My system has always had a dhclient
running and it always gets new leases when my current lease expires.  However, like any
well-administered DHCP server, that lease always contains the same IP address, because DHCP
servers keep track of which leases they hand out to which MAC addresses, and re-use them
wherever possible.  Only if my system was down for weeks and there was address pressure would
the server hand out my IP to a different host; then when my system came back up it would get
another address and all would be well.  This has worked for many years before NM came along.

There is really only one convincing use case for NM: when you need to change your network
configuration while your system is up and running.  For any system that has the same network
from bootup to shutdown (and that INCLUDES just about every DHCP environment), NM is not
needed.

Again I want to stress, I don't object to NM just because I personally don't need it, or
because I don't like new things.  I object to it because it actively BREAKS simple, basic
features that have worked just fine for years and years.  Whether that's because of NM itself,
or because it's been poorly/improperly integrated with Ubuntu and Fedora, I don't have a real
opinion about.  Whatever the reason, NM is not ready to be deployed in environments like
Ubuntu LTS.