Part of the charter? (was Re: comments ondraft-itojun-v6ops-v4mapped-harmful-00.txt)

[Alain Durand <Alain.Durand@Sun.COM>]

[ post by non-subscriber.  with the massive amount of spam, it is easy to
 miss and therefore delete mis-posts.  so fix subscription addresses! ]

Erik Nordmark wrote:

> I suspect the root of this argument is whether NAT as we know it
> in IPv4 (with DNS-ALG, FTP-ALG, etc) is "good enough".
> If so NAT-PT is what we need.
>
> Or do we want to improve on that?
> For instance, do we want improvements that allows one to take advantage
> DNSSEC through the NAT? 
> Until we agree of the right approach or approaches at that level
> I think the detail discussion is rather pointless.
I agree Itojun & I went a bit too far in the details of NAT64.
This was just to illustrate the point that IPv4-mapped addresses
can be used on the wire.

Now, to up-level the NAT discussion, the question IMHO is not
so much to ask if NAT as we know it in v4 is good enough or not,
but to really understand how v4 and v6 interoperate,
not in the early days of transition where IPv6 only exists
is some leaf sites, but when both IPv4 and IPv6 represent
a large portion of the entire Internet.

A few points:

a) The Internet is one. (RFC2826). In particular, the DNS name space is 
the same
    for both v4 and v6.
b) There will be a point where IPv6-only devices will show up.
c) There will be a point where IPv6-only networks will show up.

The current thinking for a) and b) is to heavily use dual-stack
in the infrastructure. (DNS resolver/server, SMTP relays,...)
Does this approach still makes sense for c) ?

Or another way to put it, when IPv6 penetration is < 10%,
inter-operability with IPv4 is critical, dual-stack on the
infrastructure is an acceptable price. When IPv6 penetration
is > 80%, such inter-operability may not be as important any more
But what should happen in the 20-70% range?

There are several approaches:

1) keep dual-stacking the infrastructure until the end of the times.

2) put ALGs at the boundaries of v4 and v6.
   Btw, who remember the days in Sendmail where there
   were 'well known' Bitnet and UUCP relays?
   Personal opinion, not sure I want to go there...

3) explore IP layer bridging.
   If we go that route, we can argue to see if IPv4 NAT is a
   good enough model or not, revisit NAT-PT, explore NAT64
   or invent something different.

So maybe the first question to ask is whether ot not looking
at the coexistence of IPv4 and IPv6 in this middle
phase of transition (20-70%) is part of the v6Ops charter or not.

   - Alain.