[6bone] 6bone transition

[Chuck Yerkes]

I'm catching up on mail and all of this thread.
I'm also seeing some terms being tossed around perhaps incorrectly.
I want to make some definitions and then draw the differences if
it's not clear.

6bone: An experimental backbone made up of tunnels through
       IPv4 space, direct connects and held together with chewing gum
       in some places.
       Its purpose is to allow early adopters a place to PLAY
       with and experiment with IPv6 protocols and programs.

IPv6 backbone: A set of networks running IPv6 (at least on the
       part exposed to the "backbone" that must be of production
       quality.  Like the IPv4 network and VPNs, there may be tunnels
       from one part to another across either IPv4 or IPv6.
         There may be more than one, but address really should
       never clash.  For example, my office might have a production
       IPv6 network; it MIGHT have a connection to the experimental
       6bone; strong filtering MUST be in place to protect the
       production side from flakiness of the 6bone.

To me, the term "production 6bone" is contradictory. The goal
of the 6bone (indicated by 3FFE) is that it is experimental
by nature.  If BGP erros bring everything to a stop, that's a
risk of being on the playground.  If you lose production quality
services, then that was YOUR mistake.

My use of IPv6 is entirely non-production for learning and
"getting familiar" with the tools and protocols. With OS X10.2's
release, I'll be able to drop IPv4 entirely (except for a
printer and an Annex ;), but that's the insane home LAN. Not
production.  My expectations that my tunnel be up 24x7 are
far lower than my (shattered) expectations that my PBI DSL
always be there.  Routes to sites 500 miles away are insane
on the 6bone.

I do understand that, after 6 years and some solid implementations
in OS's by default, we're ready to start rolling out some
production services.  Many of us have been working with the
protocols for years.  Many of us can recognize the value it
can present on production networks.  We're just about at the
point where we could start to have segments with no IPv4 at
all - and that can just spread virally from segment to segment
until IPv4 is remaining for some older things and tunnelled
through an IPv6 backbone.

I'd LOVE IT if some of the larger ISPs and backbones started
to feel that IPv6 was a good core transit mechanism - even if
that IPv6 part never reached the edges where the customers
join (see also OSPF use in the mid 90s).

The only way that can happen is to prepare the infrastructures
necessary to start allocating non-6bone addresses, to set up
production peering resources, etc.

While Bob's time and efforts have been invaluable to making
this experiment work and give useful information, perhaps we
might ponder giving him a rest.  I won't go wit the car accident.
I will wonder what happens when he wins the lottery and flys
off with several supermodel/masseuses/chefs, never to be seen again.

I also don't want to see the inevitable legal wrangling that
might take place with peering fall on one person's head.
(reference please Alternet's $10,000 peering fees of 1995).

The nightmare scenario, of course, is something like [i think]
ICANN.

By the time this is all worked out, other IPv6 missing bits may
be fixed (widely available NFSv6 jumps to mind).

So I view this proposal as groundwork.  Starting to clear the
lot to build new production networks that are separate from
the playgrounds that we've been using.

Not a replacement, but an enhancement.


okay, cost:  Costs will stop a lot of this from happening well.

On one hand, my site with 350 machines doesn't need and shouldn't
get a /80.  We have one connection to the (IPv4) network.  Our
collective experience is that changing prefixes is fairly
painless (change the advertised prefix, change DNS, rebind the
machines).  Certainly easier to manage than changing IPv4
addresses was in 1992.  I learned to love bootp really well
back then.

On the other hand, my previous site came THIS close to getting
a Class A, with help from Dr Cerf and nixing from Dr Postel.
We did a lot of work due to not having that.

What factors let these folks get a larger network space?

Does someone take address space away from Worldcom when they
are reduced to having 4 POPs and 2 DNS servers?

Or does their pTLA have a commodity value (as the Class B does
from a very previous employer - now with 15 employees).

Whose commodity is it and how might it be taken away from someone?
When the king of Europe is deposed and he absconds with 4FFE, by
what means can that be denied him?

But we've seen bumps here of providers disappearing, or tunnel
brokers being eaten by big companies (HP ate Compaq ate DEC).

And if one of the RIR's goes all NSI on us and gets greedy
is there recourse?  Is there an overlord who is no ICANN that
can yank their chain back?  Once the masses show up, greed and
self-destruction are soon to follow.


We've seen peering problems in the last couple months that 
bring into question peer based addresses.

Even continent based addressing is questionable.  Our asian
offices are mostly attached via VPN (or PN) to our US offices.
So our US IPv6 address might appear in the middle of Japan or
Hong Kong.  Should we have separate networks (prefixes) for
each office?

Or am I entirely overinflating the significance of this turning point?

chuck