Re: [RRG] Moving forward... IPv4 now, IPv6 less urgent and perhaps more ambitious

[Robin Whittle <rw@firstpr.com.au>]

Hi Per,

Thanks for your response.

I understand there is greater optimism about mass IPv6 adoption
amongst ISPs than I had thought.

Can you point to one or more instances where ordinary end-users are
paying for an IPv6-only service?  Or for a dual protocol IPv4 and
IPv6 service?

Has anyone done a pilot project to see how happy a real sample of
end-users would be relying on NAT-PT or whatever?

NAT-PT systems require some IPv4 space, but not one IP address for
every home customer.  Double IPv4 NAT (NATed address to the DSL
router, which has its own NAT for the customer PCs) has the same
advantage in terms of squeezing more customers onto scarce IPv4
space, without fussing with IPv6 and whatever support costs that
would entail.  However Double NAT won't work with all protocols -
and either can NAT_PT from what I know - so both are probably
unacceptable due to customer dissatisfaction and support costs.

The recent (April 200) review of approaches to NAT-PT:

http://www.arin.net/meetings/minutes/ARIN_XXI/PDF/tuesday/nat-pt-setup.pdf

points to:

  http://www.lucastomicki.net/naptd.docs.php   (Linux)
  http://www.cisco.com/en/US/docs/ios/12_2t/ipv6/SA_natpt.pdf

Linux is somewhat flaky and doesn't scale to large numbers of users.
 Cisco's approach is more stable and scalable, but poorly documented.

This small ID from Comcast:

http://tools.ietf.org/html/draft-durand-v6ops-natv4v6v4-01

discusses the pitfalls of "Double IPv4->IPv4->IPv4 NAT".  Using
"Double IPv4->IPv6->IPv4 NAT" has similar difficulties.

"IPv6 Tunneling plus carrier-grade IPv4->IPv4 NAT" doesn't help,
since "the service provider still needs to assign one IPv4 address
per customer."

That's it for the solution space.  My interpretation of this is that
none of them provide a service which an ordinary user could be happy
with and which an ISP could sell without unacceptable support costs
when some application doesn't work as it should.

It is tricky predicting how IPv4 address space could be used more
effectively, but there is clearly a strong motivation to do so,
since at present and for the foreseeable future, an IPv4 address for
each DSL customer is the only way of providing an Internet service
which meets ordinary end-user's needs without any fuss.

There may not be such a lot of scope for improving utilisation of
IPv4 space in ISP networks, since my impression is they have these
prefixes already chock-a-block with customers.

There is a great deal of scope for map-encap serving the needs of
end-user networks who need multihomeable, portable, space with less
than 256 IP addresses, so reducing the demand for conventional
BGP-managed PI space.

Maybe more applications than I realised are ready for IPv6 - at
least those being currently maintained.  But that isn't much help
for the routing scaling problem until most hosts have IPv6 and there
is no need for most end-users to have an IPv4 address.

There is still a great chicken and egg problem here, and my guess is
that the most likely outcome will be incremental - more and more
intensive use of IPv4 - rather than risky and revolutionary: trying
to sell end-user networks or individual home and office end-users an
IPv6 service which can't meet their need to use all existing
applications without fuss with all existing (IPv4) hosts.

> A safe bet on this planet is that scarcity brings out the worst in
> people. There's no reason to assume that IP will be an exception. 

It may be tempting to think of the IPv4 situation being hopeless so
the masses will soon be forced to jump ship to IPv6.  But I think
this involves getting millions of people to pay for and be happy
with a service which can't do what ordinary IPv4 Internet services
do today.


> Many
> providers that have a strategy beyond blindly copying others are already
> planning for solutions that enable growth through IPv6 at the customer
> edge. The widely anticipated steep increase in cost per IPv4 address
> will be a powerful catalyst to push investments towards more
> future-proof solutions. Regardless of an increased ability to
> slice-and-dice v4 address-blocks, nobody has yet presented a realiable
> scheme to reclaim enough v4 addresses to maintain anything that
> resembles current growth at network's customer edge past depletion.

There are 3.7 billion IPv4 addresses available.  Probably a few
hundred million of the currently advertised 1.7 billion addresses
are in use:

  http://www.isi.edu/~johnh/PAPERS/Heidemann08a.pdf
  http://www.firstpr.com.au/ip/host-density-per-prefix/

I think that provides plenty of scope for better utilization for
another decade or two, especially with map-encap.  Without
map-encap, there will be finer slicing and more BGP advertised
prefixes - the routing scaling problem heating up.  But still I
think the costs will not be so high as to make it attractive to try
to sell a second-rate problematic service to millions of people in a
competitive environment.  (In China and some other countries,
perhaps there won't be a competitive environment, so IPv6 may well
get its first widespread adoption there.)


> While IPv4 generally is expected to stay around for a long time there's
> also many people in the ops-community who believe that the majority of
> the internet may switch to v6 is as little as 3-5 years once it gets
> rolling 

Yes, but how will it get rolling?  You need millions of customers to
 adopt IPv6-only services, and be happy with them and pay for them
without lots of support calls.  Otherwise, no-one will develop the
most-used applications to the point where they are reliable and well
documented with IPv6.

How are you going to do that when IPv6 offers no advantage to the
end-user?  IPv6 is a cost-saving thing for the ISP (only to the
extent that an IPv4 address is more expensive), but it doesn't
provide what the customer really wants and needs, which is ordinary
IPv4 connectivity.

That chicken and egg thing wants to happen, but neither the chicken
nor the egg are substantial enough to bring the other into existence.

> helped by the ever shortening life-span of both software and
> hardware. Besides, it is also reasonable to expect a development in
> NA(P)T-PT/ALG solutions for v6-only hosts in the next few years that is
> similar to the rise of NAT in the late 90s.
> 
> Summary: yes we want a new (or supplemental) scalable routing
> architecture, but the v4 run-out is highly questionable as an argument
> in favour of any particular solution.

I would want to see concrete evidence of ordinary end-users being
happy with an IPv6-only service before I could imagine widespread
IPv6 adoption (home, office and corporate end-users and virtually
all web servers etc.), widespread OS and application upgrades and
significant migration from IPv4.

Only then might I be tempted to bet that half or more end-users
might be able to do without their own IPv4 address by 2015 to 2018.

Only then might I be temped to think it would be OK to let IPv4
fester without an architectural upgrade, and think only of working
on IPv6.


  - Robin


--
to unsubscribe send a message to rrg-request@psg.com with the
word 'unsubscribe' in a single line as the message text body.
archive: <http://psg.com/lists/rrg/> & ftp://psg.com/pub/lists/rrg