Re: REVIEW NEEDED: draft-ietf-v6ops-ent-analysis-00.txt

[Tim Chown <tjc@ecs.soton.ac.uk>]

On Thu, Oct 14, 2004 at 02:57:41PM +0200, Brian E Carpenter wrote:
> 
> Let me rephrase my concern. Somebody reading this draft without
> the background knowledge of this WG will simply not realise
> that the fundamental coexistence model is dual stack. They will
> find themselves right in the discussion of the 13 scenarios
> and think that they must pick one of them - but the first question
> anyone should ask is "can I just do a straightforward dual stack
> model?" I would argue that for the large majority of enterprise
> customers the answer will be yes, except for the corner cases
> where they will have to do something special. So I think the draft
> needs to start out by saying this - either by inserting my Scenario 0
> or by saying it in words.
> 
> I don't have this problem with draft-ietf-v6ops-ent-scenarios-05.txt,
> which starts out with base scenario 1, widespread dual stack.

Three points:

a) There are many deployment models, all valid ones.

b) Dual-stack is the one that most sites will follow, for the foreseeable
   future

c) IPv6 dominant has enough usage cases to be included

So I would agree with adding an ent-scenarios like wording to indicate that
dual-stack is the way to go for the more common sites, but there's more
to the analysis (IPv6 dominant was one of the three ent-scenarios scenarios
after all).

> It's not only that. Enterprises are not going to voluntarily add IPv6
> support if they perceive it as adding great operational complexity,
> which is an ongoing cost. They will simply wait until it becomes simple.

Absolutely, as per the othr email on parallel networking (as an example)
if you can upgrade that's great.  If you need to wait to a procurement
cycle, so be it, but many sites will want to jump in ahead of the commercial
availability, and that's where some of the interim methods described in
ent-analysis come in (else it is a short document :)
 
> >Not really reread the vlan discussion.
> 
> As far as I can see, that only splits the routing. As far as hosts,
> DNS and middleware is concerned it's still a dual stack model.

Yes.  And that's the advantage, but deploying parallel now we get ahead of
the game for when we can slot in dual-stack routing infrastructure, and
we already have all our services IPv6-capable (and some interesting
experience from running v6-only on the parallel routing infrastructure).
 
-- 
Tim