Re: ocean: do not boil

[Rob Austein <sra+v6ops@hactrn.net>]

At Mon, 23 Sep 2002 16:46:18 -0400, Margaret Wasserman wrote:
> 
> >What do we say to operators who can't get enough
> >v4 addreses? Deploy a v4 NAT as well as IPv6?
> 
> Why not?  How do we think that our attempts at v6->v4
> translation will be better than deploying an IPv4 NAT?

Bingo.

> >I can understand that communication between a v4only
> >host initiating a connection with a v6 only host
> >is too difficult and probably should not be a priority.
> 
> So, perhaps we are all in agreement on that.

I mostly agree.

- I think it's not too hard to add magic translating routers (or
  whatever) that let IPv4-only boxes at an edge site speak a limited
  form of IPv6 as well (been there, done that, at the detail level
  it's no worse than IPv4 NAPT and in some ways it's not as bad).

- The exactly parallel case of IPv6-only boxes at an edge site is not
  very hard either, but it's also not very interesting, since
  deploying small IPv6-only edge sites onto a dual-service network is
  not likely to be a particularly useful thing to do any time soon.

- The interesting IPv6-only translation case is a big IPv6-only cloud,
  and that one -is- hard, because the translators don't scale well.

- I think that we have enough important work to do that does not
  require translation that I think we should focus on the
  non-translation cases (see, I said I mostly agreed).