6bone - experiment into production (no longer pTLA/MICROSOFT)

[Chuck Yerkes]

Quoting -------- <removed from a private mail>
...
> Things is, like I mentioned, we may just need the marketing they can 
> provide to help people understand that ipv6 is available, and should be 
> used ...
.... (other interesting, but not relevant stuff removed)

I don't expect Windows with 6 only in the next 5 years.  I do
expect that the BACK END of large systems can become 6-only.
E.g. All the routers and infrastructure hosts in large ISP's
(Worldcom, Sprint, AOL, etc) and the parts that are hidden
from desktop machines.  So might Hotmail or the .net mess start
being 6-only?  Sure, in the infrastructure.

As these clouds of 6 grow, where they join, they can drop ip4.
So where Sprint peers with Worldcom, it may become IPv6 with
gateways at the edges.

Part of this, of course, requires that the routers themselves
be able to gateway 6-4.  Those reams of dialups that Earthlink
uses will still run IPv4 only for a LONG time.  The next hop
has to start to be able to be that transition/translation
point.  We can hold out like that for the YEARS it will take
for end-user tools to be IPv6 aware.

At some point the "6bone" has to become ancillary as "real"
production 6networks appear.  Doesn't seem like much difference.
The addresses I use are production for all intents and purposes,
just like the 10.0.0.0 and other RFC1918 addresses that are
at clients' sites.  If my application needs were met, I could
certainly run the infrastructure on 6.  I can run 6 AND 4 right
now just fine on the desktops that are ready for it.

The experimental part seems to be the interconnectedness of
any networks running IPv6.

The folks on this list are into the bleeding edge.  I'd offer
that with *BSD, Solaris, Linux, AIX, Irix and others shipping
with IPv6 for the last couple years, *and* now with Cisco
shipping IPv6 support in the routers - important glue - that
the pieces are in falling into place to have production 6
networks.  (okay, apps like NFS, portmap and such are lagging,
but I expect many of those to be fixed in the next year -
perhaps 2 years for the closed-source OS's with only limited
resources).

The point is that we're at a turning point where 6 is available
for general use.  Some tools are certainly missing, but what?

What needs to occur to take the 6 network from a playground/test
network onward into real production?

Are there companies that are willing to show that it's ready
for the real world.  Is it ready for a place like Worldcom/UUNet
to start to use it's core at least?  Is it ready for Covad to
use from the DSL connection points out to their providers?

What's the time line like for that technically?  Politically?
Are seed projects panning out?

The more pressing points to me are where there are growth
explosions.  As PDAs and phones become connected, it's not
inconceivable that CellOne or AT&T might come forward with a
need for a several million addresses.  Is the equipment ready
to have them run a standard, or do they stick with proprietary
protocols with translating proxies?

I'd love to have some 6-net discussions at the next Usenix.
It's too late (and perhaps too sparse) for a track, but a
Works in progress or BOF, at the least, could be fruitful.


Discuss.

chuck