RE: ND-proxy applicability in Unmanaged [Re: WG Last Call: draft-ietf-v6ops-unmaneval-01.txt]

["Christian Huitema" <huitema@windows.microsoft.com>]

> > Luckily enough ND-proxy is not going to be IETF standard, but
> > Informational :).  Nobody has bothered to document (different
flavors
> > of) proxy-ARP, but that's there in the similar way as well --
without
> > spanning tree, and seems to be working whereever it's used.
> 
> Sorry, I missed the last sentence.
> 
> The proxy arp implementations I know of decrement ttl.
> Are there proxy arp implementations that do not decrement ttl?
> 
> Can't compare that with ndproxy which doesn't decrement the hop count.

Actually, I know of quite a few NAT boxes that do not decrement the TTL.

I see "ndproxy" in much the same light as I see NAT: a cheap hack that
allows multiple hosts to connect to a single attachment without
explicitly requiring individual addresses or an explicit prefix. In the
same application domain, I would rather see vendors shipping an IPv6
ndproxy than an IPv6 NAT. 

Basic NDproxy only works with limited topologies. However, it does work
well with the simple "single subnet in the home" topology. In that
topology, even SEND makes sense -- at least between machines on the home
network. The combination of "works well in many cases" and "can be
deployed autonomously" is known to guarantee some deployment -- much
like the same properties allowed the deployment of NAT, even as NAT were
not described in any IETF standard. 

I agree that IETF standardization should focus on multi-link subnets,
which solves the very nasty "cascade of NAT" problems. However, there is
a lot of value in a document that says "if you were thinking of
developing an IPv6 NAT, please do NDproxy instead". The proper procedure
may in fact be a publication as proposed standard, that could then be
made obsolete by the proposed standard for multi-link subnets.

-- Christian Huitema