RE: The Network is Chaos should Routing be based on that Chaos

["Bound, Jim" <Jim.Bound@hp.com>]

>I'm sure that I need a couple of shots of Scotch before I'm in the
frame of mind
>to fully understand where you're going Jim, but lemme see if I can
clarify.

Coffee works better :---)

>You're suggesting that routing not only know where static hosts are,
but also
>ones that are changing location and that routing try to 'predict' where
the host
>will be next.  Did I get that right?

>If I could write code that did that, I'd use it on the stock market,
not on the
>Internet.  ;-)  I think you agree in your last line.  I'm not sure what
you were
>proposing at the last bit tho.  Care to expound? 

Sure.

Assume for discussion there are only two locations X and Y.

If X is location for X1, X2, ....Xn and Y is location for Y1, Y2, ...Yn


X only has to worry about its routes and only if something comes to X
for Y.

If X1 moves to Y it becomes Y-'X1' (being the global id of X but now
part of Y)

So Locations are static to start simply for high order route prefixes or
locators how ever they are to be defined.

Now suppose that Y gets to big or to chaotic and the routing
tables/performance/etc is predicted to be bad soon.  

Then we have a way to take Y and make Y and Z and now we have three
locations X, Y, and Z.  THis would be a formal location change control
routing protocol.

So I am not saying predict movement but design it into the algorithm for
location "expansion" via routing protoocol and inherent to the
architecture.  

Location change would have to be propogated across the system yes.

Using ipv6mh ideas if we define specific locations this can be done for
a long time with IPv6 128bits on planet earth.  That's a lot of bits.

So this is a means to adapt to the chaos of change we cannot see.  Like
right now and more so very soon imagine all the IP devices about to
descend.  Then nano sensors.  Then stuff I can't imagine right now
without some sleep :---) THe point is the system keeps growing and then
splitting much like our galaxy :--) There is no order its all chaos.
All we can do is see patterns and adapt through location routing
updates.

That is why extensibility is key to any protocol (earlier comment this
a.m.).  We reached the limit with IPv4 and now bad stuff is going to
happen if it is not replaced expediently. So if we can compromise here
on an answer with IPv6 and not break apps/apis/basic stuff but alter to
get to location + identifier we are probably ok for 20-30 years.  And if
not well we do it again.  Not much lasts longer than that regardless of
what we do.  The IETF is an infinite recursive effort for all time now
:---)  Unless we really screw up and blow up the planet.

Did that help?

/jim