Re: [Autoconf] answers on DHCP DUID, key generation etc.

["Charles E. Perkins" <charles.perkins at earthlink.net>]

Hello Fred,

Comments below...

Templin, Fred L wrote:
> > Yes.  Obviously DHCP has a strong role in the administration
> > of fixed networks and even infrastructure-based networks with
> > mobile nodes.
> >     
>
> DHCP is for enterprise networks in which there is some
> modicum of distributed network administration and a
> commonality of interests (which may be as basic as the
> collaborative provisioning of connectivity itself). This
> model fits well not only with infrastructure-based
> networks but also with many MANET scenarios with which
> I am familiar.
>   

I agree there are wireless mobile networks that support
DHCP.  It's nicer, however, when the server and the relays
have stable availability and connectivity.

>  
>   
> > > I think DHCP can be very useful for attached MANETs, for getting
> > > "additional information", i.e. other than addresses.
> > >
> > >       
> > I agree.  I just don't think it fits very well in networks composed
> > of highly mobile wireless nodes, especially when the network has
> > no point of attachment to the Internet.
> >     
>
> Like maybe me and a couple of buddies going to a remote
> ski area with hand-held multi-hop walkie-talkies when
> there is no cellular coverage? Even then, unless we all
> get together beforehand and key our L2 access codes what
> we get is "anarchy-net", and anyone can crash the party.
>   

As you know, even enterprise DHCP had security problems
for a really long time.

There are many flavors of ad hoc networks between the ski-slope
anarchy-net and the attached wireless stubs most conducive
to supporting DHCP.  My main point is that we can't just wave
the DHCP wand and declare mission accomplished.

But as an interesting point of distraction, consider: (a) mutual
face-to-face setup of security association (cellphones are cool)
and (b) terabytes of storage in your snowboard.

> This brings up again a point which I have made in the past 
> that this group could significantly benefit from a detailed
> MANET use-case analysis before saying what the addressing
> architecture for *all* MANETs ought to look like.
>   

But nobody is doing that, Fred.  Right now we have just about
the most minimalistic document you could imagine.  Other more
specialized applications should be derivable from the general model
hypothesizing appropriate characterizations for the more specialized
model.

For instance, if we have a system that works when there aren't
any guaranteed unique MAC addresses, then that system CAN
be specialized to the case of unique MAC addresses.

The other way around, is often impossible.  It is unlikely that you
would be able to generalize a guaranteed-unique MAC address
solution to fit the needs of systems that do not guarantee such uniqueness.

Frankly, I think this should be considered very obvious.
Too bad we've been bogged down for years now explaining
it over and over again.  It's like claiming that we cannot
go about designing solutions for x+y = 4 because a lot of
times x is equal to 0.

Regards,
Charlie P.