[rrg] solution strategies, nat?

[Steven Blake]

On Wed, 2008-12-31 at 11:57 -0500, William Herrin wrote:

> On Wed, Dec 31, 2008 at 11:07 AM, RJ Atkinson <rja at extremenetworks.com> wrote:
> > I agree with Steve Blake that a NAT-oriented approach,
> > which I gather is category "H", ought to get serious consideration.
> 
> Ran,
> 
> Would you care to describe a NAT category which credibly addresses the
> routing scalability problem in a manner not already encompassed in
> widely deployed NAT systems?

That's the point: widely deployed NAT (as it exists today) is a
do-nothing strategy from an interdomain routing point-of-view.

> I have no intention of adding:
> 
> Strategy H: We should use NAT!
> Major criticisms: We should use NAT to do what exactly?

Allow sites to have stable internal addressing without putting pressure
on interdomain routing scalability from additional PA prefixes.

PROS: 

- Unilateral deployment incentives: no need for any global
  coordination, or involvement by the providers.

- Widely tested, mature technology (for IPv4, at least).

CONS:

- Does not provide session survivability in the event of uplink failover
  (at least not with current IPv4/IPv6 and transport protocols).

- N:1 address multiplexing (v4 NAT) makes it difficult to support
  inbound connections.  v6 NAT could be better in this respect.

- Breaks certain classes of applications (i.e., those that depend on
  address referrals).  In some environments this is considered a
  feature, not a bug.

All of these cons could be mitigated by evolution of IP and/or transport
protocols.  As long as these enhancements support backwards
compatibility, they could be incrementally deployed.  Many have already
been discussed in RRG.

PS. Nobody despises NAT more than I do.


Regards,

// Steve