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

[Andrew Sullivan]

Hi,

I don't think this thread is going anywhere, and I think there is now
ample technical evidence that what you're talking about is unworkable,
so this will be my last posting on this topic.

On Thu, May 22, 2014 at 03:56:53PM -0600, willi uebelherr wrote:
> But this is very simple. You change the DNS entry for your
> home-address. Then all people have your new IP-address. And if you
> are not able to do it, you can ask friends or neighbors to do it for
> you.

So, just so we're clear:

Today, on the actual Internet we have, if I am the registrant of an IP
address range and I move my data centre from one location to another,
I make a new announcement and everyone can find me automatically.  

In the new network that you envision, if I have a set of services on
the geographic-based network and I move my data centre, I have to
update all of my software configuration to tell my various servers
about the new address and I have to update the DNS to tell everyone
where to reach me.  Probably, to make this work properly, I have to
update the DNS before and after doing all this, too, so that old
values aren't cached for too long (and then to restore the benefits of
caching).

I confess that this seems to me like a lot of extra needless work in
order to achieve something of dubious value.  Indeed, the trend in
networking is _away_ from the location/identity link, not towards it.
Perhaps you're in possession of some piece of information I don't
have, but I think you're siezed on a network topology that has been
proven not to be as convenient to its users as some alternatives.
Particularly in a mobile world, where many (most?) nodes are mobile
all the time.

> Agreements, yes. Based on the same intentions and interests for
> transport of data packets. Based on the using of the same IP header
> structure. Based on regional activity for the interconnection of
> regional centers. But "Governance"? One or more institutions or
> organisations that organize it as a representative instance? No.

I think that some of us have been arguing, at length, on this very
list that the distinction you are drawing above is _exactly_ the one
that obtains on the Internet today -- i.e. that many of the
"governance" issues that people are worried about are either attempts
to create a problem so that the would-be governing party gets control
it doesn't already have, or else the upshot of a misapplication of
rules of scarcity based on a misuse of the underlying technology, or
else some things about which there are genuine disputes.  Given that
you're now talking about radio spectrum -- a resource that is, as a
matter of fact, naturally scarce -- it is hard for me to see how your
solution is going to avoid the genuine disputes.  We can't even avoid
disputes about allocation across wires or fibre-optic cable (which
could be made less scarce by laying more).

> "It still requires dispute resolution if two diffrent people are
> stepping on the same location (because, for instance, two ISPs are
> in racks next to each other in the same data centre -- i.e. below
> the resolution of GPS)."
> 
> No. In my last answer to you from 17.05.2014 you see the resolution
> for the 64 bit global address part.

That only works if you can resolve the global _and local_ address
parts in such a way that nothing collides.  You've simply waved away
that problem.  Indeed,

> But what have this to do with the GPS-system?

the GPS system is the only actually deployed system of which I am
aware that would tell you the location of all these nodes.  The nodes
they might move in between packets, because we have an Internet that
includes mobile devices.  Perhaps with your idea that people don't
often communicate outside their geographic communities, you forgot
about mobile devices? 

> And ISP's? We never use it. We organize the interconnection of the
> local networks self. Only for the intercontinental connection we
> need a deeper cooperation with the people in the other parts of our
> world. Like the people in Asia, Africa and Latin America.

There are two possibilities here.  Either geographically local
connections are somehow aggregated towards the intercontinental
connection points, or they're not.  If they _are_ aggregated (because
for instance you talk to a larger antenna, that then talks to some
other antenna and so on) then you need something analagous to an ISP.
If they're not aggregated, then you're supposing that every mobile
phone is going to have a perfect antenna that can communicate over
long distances.  I look forward to your calculations about battery
life.  Also, I invite you to go to a very busy meeting in a hotel and
see how well the radios cope with so many transceivers.

I appreciate the goals you are trying to promote, and I think if you
do a little reading you will find that some of the early premises of
internetworking started from similar premises.  The difference is that
the people working in that case understood the disadvantages of
strictly geographically-bound network topology, and attempted to deal
with that.  You seem not to have considered all that prior work.
Moreover, there _are_ people who are working on a future network
architecture that might tackle some of these issues differently than
the network we have today; consider "Named Data Networking", for
instance (http://named-data.net/).  (Lest anyone get too exited: if
you think problems about name conflict are bad today, wait until
trademark lawyers can disrupt the routing system.)  But linking
Internet routing tightly to geography would be a quantum leap
backwards.  It causes large practical problems and I see no benefit it
offers.

Best regards,

A

-- 
Andrew Sullivan
ajs at anvilwalrusden.com