[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Dominant IPv6 Network deployment for Transition by Users

On 23-jun-04, at 23:09, Pekka Savola wrote:

I think you assume that having to support one dominant protocol (IPv6)
and one less dominant (but still a MUST work protocol, IPv4), through
mechanisms more complex than just deploying IPv4 (or keeping it
deployed) requires a smaller amount of support in total?

This seems dubious to me.

There are many cases where IPv4 is quite troublesome. For instance, if you organize an event where lots of ad-hoc users show up, you need a lot of IPv4 address space that's hard to get if you don't already have it. And any time you build a network that has significant growth, you end up with very fragmented address usage, which in turn can make it necessary to have huge broadcast/flooding domains with all the trouble that entails. With IPv6, none of this is an issue, so it's much easier to design a network and then deploy it without having to change much later. Now obviously supporting v4 connectivity over tunnels brings back some of the v4 complexity, but I expect there will still be significant advantages in many deployments.

Note though that I wouldn't expect each and every host to have an IPv4 address in such a scenario: hosts that only use simple client/server protocols can either use address translation or proxies.

3. It is far easier to control the operation of transition to IPv6 once
IPv6 networks are dominant and IPv4 is treated as legacy. 

Do we (and the customers, or at least the majority of them) actually
want to control the operation of transition to IPv6-only at this
point?

That's not the point Jim was making. If you want to run both IPv4 and IPv6 there are several ways to do this: run v4 first and add v6 where needed (this is how it's done most of the time today), run them as equals side by side, or build an infrastructure that's optimized for IPv6 and then bring back IPv4 where needed the same way legacy protocols are generally supported. Whether IPv4 is actually legacy or not is pretty much an eye of the beholder thing, what counts is the way you design a network. Building a significant NAT-less IPv4 network is very challenging, much more so than building an equivalent IPv6 network. Then adding back IPv4 through tunneling has the advantage that the address usage is no longer tied to the network topology so it's much easier to meet RIR policy requirements.

I imagine most would want to deploy IPv6 because it brings
them a benefit they want.

Simple network design and operation is what many people want. :-)

Until a significant portion of the Internet
has adopted IPv6, an easy strategy could be to postpone the decision
on when to move to IPv6-only.

Sure. But note that enterprise users usually talk much more amongst themselves than with the internet at large. The issues are different from ISP networks or home networks, where connectivity to the rest of the world is the most important thing.

Also note that this isn't a question of running IPv6-only hosts, the hosts will still be dual stack. It's a question of how to provide dual stack connectivity to a large number of hosts in the way that's most efficient.