RE: PI/metro/geo [Re: The state of IPv6 multihoming development]

["Tony Hain" <alh-ietf@tndh.net>]

Tony Li wrote:
> ...
> |   That question deserves a firsthand answer from a current 
> enterprise
> |   network manager.
> 
> 
> So it may surprise you to learn that I used to be a network operator
> some time ago [Los Nettos, 1987-1990] and an enterprise 
> network manager
> for my company and of course my own home.   

That does not surprise me, most people that really understand
inter-domain routing have arrived from the school of hard knocks. The
point I was trying to make is that the reasons behind deployments today
might be different from those we used 10+ years ago, so we could use a
first hand answer to current demands.

> We happened to recently change
> providers at my company and, because we are behind a NAT system, found
> the change to be quite trivial.  Of course, we only needed to 
> renumber the
> NAT box.

That will work for some companies with limited application needs, but
for others nat is a non-starter. 

> 
> In a 16+16 environment, where only the border routers are 
> configured with
> the global locators for the enterprise, renumbering amounts 
> to reconfiguring
> only one system.

Well, for the general case that one 'system' needs to include the border
router, dns, firewall, ids, partner access lists, etc...

> 
> 
> |   But since there is no way to ensure that an identifier is globally
> |   unique, the only way to accomplish the goal is to couple them to
> |   describe 'this identifier at this location'. If the 
> location is not
> |   stable, the static description is not stable. 
> 
> 
> You can assign an identifier based purely on the 
> administrative hierarchy.
> We have an existance proof that we can do this, because we've done so 
> already in DNS.

In fact we have an existence proof in both DNS & IEEE EUI that
inadvertent & intentional duplication happens. So those mechanisms can't
be used as 'globally unique' identifiers as they are. If we add some
cryptographic properties, we can probably improve that.

> ...
> I submit to you that if folks are unwilling to change to an
> architecture that scales, then IPv6 is already doomed.

If the changes are contained to the routers, the timeframe is reduced
from 7 years to 3. In any case a drastic change will take two years to
get vendor shipments and another year for service providers to deploy. 

> 
> 
> |   Because the identifier is not stable & globally unique, and 
> |   even if we
> |   had a way to ensure that, the privacy advocate's alarms 
> |   would be going
> |   off. Despite the desire to avoid it, the current reality 
> is that to
> |   create a globally unique identifier we bound the problem by 
> |   coupling it
> |   with its locator.
> 
> 
> That is certainly true today.  However, there is no requirement for
> that.  Consider that host foo.cisco.com already has a globally 
> unique and fixed hostname.  We allocated this in a hierarchical
> fashion that requires no aggregation.  We simply do the same for
> the identifier space.

That is the same argument that was used to structure the locator part
according to provider hierarchy ... ;)

That aside, until we get DNSsec widely deployed there is nothing that
prevents someone from injecting foo.cisco.com into a remote part of the
tree. The only way we can consider a bit string to be a globally unique
identifier would be to have some cryptographic property that could be
checked in real time. I am not opposed to such a system being created,
but holding IPv6 multihoming hostage until it exists would be a
disservice to those who need expanded multihomed address space now.

Tony