Dan Lanciani wrote:
>
> "Tony Hain" <[EMAIL PROTECTED]> wrote:
>
> |That said, there are multiple parts to the isolation issue, and even though
> |most NAT implementations combine them, the discussion will be more
> |constructive to keep them separate.
>
> I used to believe this, but I recently came to the realization that isolation
> from provider address policy is a single issue.
Perhaps, but it is not an issue to which NAT is the only answer. In IPv6 we have
enough address space to solve it without NAT.
This is one more round in a circular argument. Yes, there is more than ample address space, but there is a common perception that there is not ample routing space.
It seems to me that there are a set of very loosely coordinated efforts in this space, that are all, ultimately, directed at the same issue - attempting to avoid placing fatal pressure of the routing system.
On the one hand there is an effort to constrain the supply of discretely routeable elements through: policy constraints on the allocation of provider independent address space, use of "private" address space coupled with public NATs, consideration of various forms of scoped address architecture, and use of tunnelling and encapsulation.
At the same time we're attempting to drive an effort to improve the scale of the routing system through a number of research related efforts that attempt, in various ways, to scale routing by looking at the routing domain as a synchronized discovery of topology, policy and address prefix reachability.
At the same time we see vendors making continual improvements in router architectures and their components that lift the practical ceiling of routing capability at a scale that appears to continue to track a reasonable approximation of Moore's Law.
Coupled with continual improvement in operational environments that have, on the whole, created improvements in the stability of the routing environment.
I have the _personal_ impression that we've managed to overdamp the supply of address prefixes such that the current routing system is now situated very comfortably within the current capabilities of the routing system and the underlying active network elements, but at the cost of increasing application complexity in a world of NATs, dual protocol stacks, multi-homing, mobility, and so on. Its not entirely clear to me that this approach scales. My impression is that we've not managed to push the scaleability of routing technology at the same time, so we've fixated, perhaps too far, on applying constraints within the address system and paying the cost through increasing application and service architecture complexity.
If we had more confidence in the ability to scale the routing system we could look beyond the relative merits of various forms of application of address system constraints and strike a better balance between carrying the load of further network scaling between the routing and addressing systems.
Geoff
-------------------------------------------------------------------- IETF IPv6 working group mailing list [EMAIL PROTECTED] Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------