RE: ocean: do not boil

[Pekka Savola <pekkas@netcore.fi>]

Some of my own comments.

On Sat, 21 Sep 2002, Bound, Jim wrote:
> > i don't think we should try to solve the problem of an arbitrary
> > user/service on a pure v6 site/host trying to communicate with a
> > user/service on a pure v4 site/host or vice versa.
> 
> Nor do I for the general case.  The market will solve this problem quickly anyway IMO. 
> The one deprecated case that the scenarios will see (all of them) is when they 
> are going to a v6site and that ASP site can only provide that service by going 
> to a v4 site. It may beg the question do we need an ASP set of scenarios in addition
> to ISP scenarios?  Which came up at the Interim meeting.

ASP is an enterprise, IMO.

If they don't feel providing v6 service important enough, you can always:
 1) switch ASP's so make the market decide
 2) implement some v6 -> v4 translation yourself (we're doing something 
remotely similar ourselves)

> > if i want my web site seen by both v4 and v6 users, then i can
> > connect it to v4 and v6 space and run dual stack.  in fact, i do so
> > today.  as v6 deploys, folk every useful site will do so.
> 
> we do this today too on the 6bone and internal net.  I don't see a pure 
> v6 site that don't have a dual stack for a long time IMO.

Yes.

And btw, in especially in enterprise scenarios, one has to really careful
about 'ipv6-only'; there are some services that usually need to be
dual-stack, like authoritative DNS servers, WWW caching proxies, and some
MX's; this can be partially worked around of course.

> > trying to make pure v4 sites/hosts communicate for arbitrary
> > services with pure v6 sites/hosts and vice verse is a nice way to
> > make the problem vastly more complicated, the solutions vastly more
> > complex, and the net much less reliable.  let's not go there.
> 
> I think I agree but not sure.  Let me provide a scenario I am working for deployment 
> as we talk.
> 
> user has v6site except for parts suppliers have not completed v6 services 
> and for now in the site uses private v4 addreses to get to them. but the v6site
> must get data from v4 site to use for its operations that requires use of global
> v4 address over Internet network.  Today the user just uses one of their global
> v4 addresses for that data transfer. But the network ops design evolution requires 
> all sites to move to dominant v6 site.  What will happen though is the network 
> edge will move last and remain v4.  The plan is to tunnel v6 inside v4 to communicate
> between two v6 sites.
> 
> The above assumes I think what you say is way to complex and the net ops plan above
> avoids that complexity.
> 
> Am I on same wave length as your vector issue?

I believe "tunneling v4 through v6" in a completely v6-enabled site is not
too complex.  If it's too complex, the decision to move to v6-only has
possibly been made prematurely (for some part of the network, at least).

In the above case, depending on the service I'd probably provide a TCP 
relay access to the service (if transport-agnostic), or just keep v4 
service until there is no more need for it.

Restricted use of TCP relay is IMO a rather good transition solution for 
some services.

-- 
Pekka Savola                 "Tell me of difficulties surmounted,
Netcore Oy                   not those you stumble over and fall"
Systems. Networks. Security.  -- Robert Jordan: A Crown of Swords