Re: Limited Domains:

[Toerless Eckert <[e-mail]>]

On Wed, Apr 14, 2021 at 11:10:54PM +0100, Nick Hilliard wrote:
> Toerless Eckert wrote on 14/04/2021 23:01:
> > Which IMHO does not exclude the option to design such a protocol for
> > limited domain in a way that it also supports "non-limited domains".
> 
> we already have this with MPLS, but it was designed from the outset as a
> limited domain protocol and was only later extended to interoperate outside
> the domain boundary, and only then in specific and carefully controlled
> circumstances.

Indeed. And this could probably improved.

> Re-scoping an existing inter-domain protocol like ipv6 into a limited domain
> situation is a fundamentally different proposition.  Unless the scoping can
> be managed on a mandatory per-packet basis, it's the sort of proposition
> that will end up causing a long wake of operational problems and
> consequences.

True. But i think we can collect the experiences we have, especially from
rfc1918 (multiple instances), ULA, IPv6 PD addressing /multiohoming and construct
from them the principles for a better protocol framework. Of course also
aspects such as the (now dead) site local addressing and (still alive)
4 bit of scope addressing in IPv6 multicast.

If we just stay where we are now, i am afraid that a lot
of industrial use of IP, which is IPv4 only will never move to IPv6
because the longer addresses and ULA are not really an improvement
over RFC1918 + NAT in the mindset of such operators. rfc1918 provides
"generic addressing", ULA not (for example). IPv4 addresses can be
memorized by operators, IPv6 addresses not (unless you use illegal
prefixes). Etc. pp.

> 
> Nick

-- 
---
[e-mail]