[address-policy-wg] RE: gazing into the future

> First, I find it staggering that anyone is contemplating -- 
> even in passing -- a replacement for IPv6 and a shopping list 
> of features for that when the level of IPv6 uptake is so low.

Some of us are dreamers and wonder what could be done if we
could just sweep away the old. If you look at history, sweeping
away the old does happen from time to time. Stroll down London's
Regent Street and there is not a trace of the squalid slum
that was there before the Prince Regent swept it away. On most
Central London streets you can still see the traces of the
medieval city in the layout and size of the buildings. But
not on Regent Street.

There are some serious researchers also doing work on
redesigning the Internet from scratch. 

Perhaps the reason for doing this is BECAUSE the uptake of IPv6
is currently so low, and BECAUSE we realize that many of the 
reasons for this is that IPv6 ideas were tacked onto the IPv4
network and therefore we won't get the benefit of those ideas
from upgrading. This lays bare all the inadequacies of IPv6
and people naturally look for ways to remove those inadequacies
but replacing IPv6 is not currently feasible.

Also, it is likely that a new generation of engineers without
the baggage of IPv4, will be better able to make a fresh start
on IPv9.

> It's unwise to assume there is no risk of IPv6 addresses 
> running out.  

That's what I said. Since there is a risk of it running out
in 100 years or so, we should plan to create a better
replacement for it in 30-50 years, and avoid the whole issue.

> The same mistake was made ~25 years ago when ARPAnet 
> transitioned from NCP to IP. A 32-bit address space was then 
> considered too big to fill.  

The arithmetic was rather different in those days. You really
should read Geoff Huston's. One source of these numbers is
this ARIN presentation
<https://www.arin.net/participate/meetings/reports/ARIN_XV/PDF/wed/Round
_Table.pdf>

We really should ask Geoff to revisit the numbers in light
of 6RD and the larger allocations that might happen. I get the
sense that this would not cause a problem as long as we assume
that 6RD will be decommissioned within 5 to 10 years. There is
no point in discussing possible runout without some actual
figures in the discussion because IPv4 and IPv6 are several
orders of magnitude different. It becomes obvious when you
work through the figures.

> Remember too that since each customer will in all probability get a /
> 64 from their ISP,

No, a /56 or /48

> the effective size of the IPv6 address 
> space in the core of the network is 64 bits. 

IPv6 addresses are 128 bits everywhere. The core uses full
addresses.


> BTW if your claims are correct, this presumes the IETF can 
> complete the development of IPv9 in 20 years. Which seems... well...  
> optimistic. :-)

I never mentioned IETF. In any case, the people who do the work
30 years from now will not be the same people who did the IPv6 
work.

--Michael Dillon