ietf
[Top] [All Lists]

Re: Death of the Internet - details at 11

2004-01-13 04:31:39
On Tue, Jan 13, 2004 at 10:30:05AM +0000, Tim Chown wrote:

It is prevalent wherever there is broadband.  And that is where (with the
extra bandwidth and always-on) connectivity into the network is desirable.

Not around me it isn't. In the UK, even with cable modem providers, I have 
non-NAT - as they are known in the European ISP industry "RIPE addresses" - 
and although I've installed NAT myself to enable quick and easy WiFi access 
using the one IP address, there is nothing stopping me taking that box out 
and having proper IPv4 direct to my NIC.

In addition, many, many broadband ISPs in the UK will not only provide IP 
addresses, but will happily route subnets providing you fill in the form 
explaing why you want the addresses, so they can give the justification to 
RIPE if they need to apply for more address space.

There are a few big players who do NAT - BT OpenWorld is the biggest I know 
- and yes, they are aiming for a consumer market that as yet does not see 
the advantage of non-NAT network access. There are two ways people will 
address the issue when the time comes:

1. Demand non-NAT network access
2. Use protocols that work around NAT restrictions
  
IPv4+NAT will coexist with IPv6 for many years.   A home router can easily
offer v4/NAT and v6 together.   This allows v6 apps to be used opportunisticly
between homes or other networks that would otherwise have NAT and need some
3rd party broker.

I am not advocating a policy of insisting that IPv4+NAT gets withdrawn from
the Internet. What I am suggesting is that there is an official withdrawal
of support for continuing the development of IPv4 protocols - that we should
be saying to the world "Look, very nice, but we're about 3 years ahead of
you on the technology curve, and we're not providing life support for
something we know you won't want in 3 years time when we can be using 
that time dreaming up the support for the apps we know you will want"...
  
That's rather insane :)   More like July 31st 2025 before we remove IPv4,
and even then it'll hang around... remember no-one *has* to install IPv6,
it's just an option if you want the functionality.   Users want features not
protocols.

I'm not saying we remove IPv4. Just we push market demands by removing
continued RFC support for protocols replying on it. This is exactly the
technique the author of Speak Freely used to bring this issue to our
attention. It's what MS, Sun, Cisco, everybody does. If you're happy with
IPv4, fine, keep it, but we're not going to carry on pumping resources into
something that quite frankly, is not a suitable use of our time. This to me 
seems a reasonable approach.

I think if we say "From the middle of next year, no more IPv4 RFCs or drafts 
please", then vendors and application developers will have to sit up and 
take notice. Remember, the protocols take between 6-36 months to be deployed 
for real, so what we'd actually be saying is "we don't think IPv4 is worth 
deploying from scratch after the middle of 2007". We'd be saying to 
application developers "Look, IPv4 isn't where you're going to make serious 
cash with innovative applications in the future, come play with IPv6".

And of course, if after a couple of years it isn't working, there is nothing 
stopping the IETF rescinding, and supporting IPv4 once more due to "customer 
pressures". :-)

-- 
Paul Robinson