dominant v6 assumptions [RE: REVIEW NEEDED: draft-ietf-v6ops-ent-analysis-00.txt (fwd)]

[Pekka Savola <pekkas@netcore.fi>]

On Tue, 26 Oct 2004, Bound, Jim wrote:
> Good comments and good response.  One important point to note that was
> noted to Pekka in my response. No one is assuming an IPv6 ONLY anything.
> The entire spec assumes dual stack.  What we do believe is users will
> move services and routing to IPv6 and not IPv4 on points of the network
> in several cases immediately as part of the transition and definitely as
> a plan.  This does not imply IPv6 only but dominant use of IPv6.

Different people have different assumptions about 'dominant IPv6', 
which might be confusing here.  For example, a couple of 
possibilities:

 1) both v4 and v6 routing and addressing is provided, but most 
services have been made available using v6, v6 ALGs have been 
configured (e.g., web proxies), so consequently, the majority of the 
traffic is going as v6.

 2) both v4 and v6 routing and addressing is provided, but there are 
some nodes which are v6-only

 3) in some places in the network, v4 routing/addressing has been 
disabled.  This case must somehow answer the question how hosts or 
routers in those places in the network will talk to v4 nodes if they 
need to (or is it assumed they won't need to?)

 4) v4 addressing/routing has been disabled everywhere except for a 
couple of boxes near the border.  See 3).

In other words, even if the nodes are dual-stack capable, if you 
switch off v4 routing or addressing, you'll have a couple things to 
think about:
 a) is it worth it to turn v4 off, would it be easier if you didn't? 
(in very many cases, it seems that turning v4 off prematurely results 
in highly increased complexity in the network)

 b) if you do, can you make the assumption that those parts of 
networks don't need to talk to v4-only hosts/routers? (then it would 
be OK to turn it off IMHO)

 c) if communication with v4 nodes/apps is still needed, will you 
perform tunneling or translation.  If translation, is it 
application-specific (e.g., ALG/proxy; that's OK) or generic (oh no, 
NAT-PT, see argument a)?  If tunneling, couldn't the hosts always have 
an IPv4 address, but just over the tunnel (also see argument a) -- 
what would be the point in that)?

--
Pekka Savola                 "You each name yourselves king, yet the
Netcore Oy                    kingdom bleeds."
Systems. Networks. Security. -- George R.R. Martin: A Clash of Kings