[discuss] Internet: the INTER-connection of local NET-works

[Andrew Sullivan]

Hello,

On Sun, May 04, 2014 at 01:00:13AM -0600, willi uebelherr wrote:
> 
> Internet: the INTER-connection of local NET-works

With one exception, I agree with the above statement.  I don't know
what "local" means.  In networking, it usually means "topologically
local", in which case "local network" and "network" are just synonyms,
and we just add the "local" to differentiate cases when we also have
more than one such local network in play (i.e. in an internetworking
case).  From other things you say, however, I get the feeling that you
mean "geographic location" when you talk about "local".  In that case,
the above statement is certainly false today.

You're nevertheless right that the Internet is the inter-networking of
all networks.  This gets to a central fact, too often overlooked in
our discussions, that the Internet is _by definition_ voluntary
carrying of one another's packets.

In any case, there is a central flaw in your proposal that makes it
unworkable in two ways.

> 3) The IP address
> 
> The IP address is derived from the geographical position in the
> world coordinate system. We use 64-bit for global and 64-bit for
> local address. Because the world coordinate system WK84 is
> distributed asymmetrically, we should strive for a symmetrical
> system of coordinates. Maybe it already exists.

Given the other things you say, it seems to me that your proposal
requires that the addresses are _strictly_ linked to geography,
because routing has to use the geographic location of a device in
order to route packets.  That's going to be really hard to build a
network atop, because if the end point is in motion its address
changes all the time.  Indeed, its address might change for every
packet, and I would have to know in advance where the other end of my
connection was going to be just to send the packet.  Given that an
enormous amount of the growth in traffic comes from mobile devices,
this is not a theoretical problem.

To make this even logically possible, therefore, you need to invent a
protocol by which I can tell everyone else on the Internet where I am,
apparently without any central co-ordinating function).  If the answer
is "GPS", it won't do, because the GPS resolution is insufficiently
fine to differentiate two devices that I am carrying with me.

If you're going to argue that you can use GPS to co-ordinate router
addresses and then let the router sort out the addresses it routes,
that's still no use.  You could have more than one router in the same
rack, for instance, and GPS resolution isn't fine enough for this
(even supposing that we got to use military precision, which seems
far-fetched).  You also then have a problem of address governance,
it's just distributed, because there needs to be some way by which a
router decides which address is associated with which device, and
we're back into the same ball game we're playing today.

So, quite apart from the technical problems that were caused by
nailing identifiers to physical location (a problem bad enough that
phone companies did work and spent money many years ago to break that
tight relationship, some time before mobile devices were important),
as near as I can tell your proposal is not technically feasible at
all.

Supposing it were feasible, however, you're setting us up for a
different problem.  By linking all of this stuff to geography, you're
throwing away an enormous number of addresses.  You have the first 64
bits for routers, but that's based on location.  Since routers are
unlikely to be uniformly distributed around the Earth, the bulk of all
address space will be empty.  Also, some places will be very dense
with devices.  Maybe 64 bits is plenty for all the addresses in a
given router's geolocal area, but at the very least you're going to
need to do the analysis.  (In passing, also, I note that you had
"regional" things in your description, but there didn't seem to be a
way of identifying regions.  It strikes me anyway that regions are not
obviously politically neutral, so I have a hard time seeing how
they're going to get us away from the problems of governance.)

> The routing (geo-routing) is based on the destination address of the
> packet relative to the position of the router. From the distance and
> the angle wc can easy make the decisions.

This empahsis on distance and angle certainly explains why you wanted
to specify the link technology in your addressing architecture.  That
seems to me like an excellent way to bake in obsolete technology,
however.  What if we're not using radio waves?

Also, note, that your addressing system works only for Earth-bound
systems.  You need a completely new addressing scheme for (for
instance) networking with spacecraft.  Since work has already started
on interplanetary internetworking (see
e.g. http://www.nasa.gov/mission_pages/station/research/experiments/730.html),
this is going to be a problem.

Finally, and perhaps most importantly, I can't see how any of your
proposal can interoperate with the existing Internet.  If it can't,
then what you are suggesting is that we have to cut over to some
completely new architecture.  Presumably, we'd have to do that all at
once, or else the intermediate co-ordination mechanism would have a
very long life (and getting rid of such mechanisms is the whole point
of your proposal, AFAICT).  So that requires a "flag day".  The last
time we had a co-ordinated switchover of everything on the Internet,
you could get a list, on paper, of the names and addresses of all the
people (not endpoints!) connected to the Internet.  That event didn't
come off as planned.  Why do you think a co-ordinated switchover would
work this time?  If you don't think you need one, please explain why
not.

Best regards,

A

-- 
Andrew Sullivan
ajs at anvilwalrusden.com