ietf
[Top] [All Lists]

Re: A simple question

2003-04-29 06:27:22
    Date:        Mon, 28 Apr 2003 05:53:20 -0400
    From:        Margaret Wasserman <mrw(_at_)windriver(_dot_)com>
    Message-ID:  
<5(_dot_)1(_dot_)0(_dot_)14(_dot_)2(_dot_)20030428053937(_dot_)047cbc68(_at_)mail(_dot_)windriver(_dot_)com>

  | In particular, you would still need to renumber your local network
  | (the global prefixes) when your provider-allocated global
  | addresses change.

Yes.

  | Having extra addresses available for internal
  | traffic (the site-locals) does not make renumbering the global
  | prefix any easier or less expensive.

No, that's simply wrong.   One of the biggest costs in renumbering
is the disruption it causes.   The actual cost of editing the files,
etc, is trivial by comparison.

If all of the internal operations of an organisation are going to
be disrupted, then the organisation is really going to resist any
renumbering.

But if everything just sails on through, operating as normal, then
adding a prefix, and later deleting an old one, is must less intrusive,
and less costly.

  | Although NAT causes various problems, it does offer a high degree
  | of provider-independence for internal nodes.  You won't get this
  | using provider-allocated global addresses in IPv6, no matter how
  | many other addresses you add to each node.

It would be nice if everyone would stop using "provider" in this context.

What matters here is the use of topology sensitive addresses.   If the
address reflects the topology of the network, then it has to change when
the topology changes.

It happens that in the current network environment, the topology is
controlled by the providers, so we have provider supplied addressing
to correctly reflect the topology.

Getting provider independent addressing is plausible - we just need
to make the network topology (somehow) stop being controlled by the
providers.

Getting addressing that is independent of the topology is a much more
interesting problem.  That one I don't believe we have any way to
accomplish yet, that works with routing.   Until we do, we need
addressing for local use.

  | Of course, this isn't why NAT is most often used...  NAT is most
  | often used to extend a single address to cover multiple systems
  | in a home or small office environment.

I think you're living in a sanitised enviornment.   NAT is used almost
everywhere (perhaps outside the US) - almost everywhere.

  | For that environment,
  | an IPv6 /48 (without site-locals) would suffice to replace NAT.

It might.   I'm not sure it will however.   I don't know that I want
even my home connectivity (connections between controllers and
appliances and such) to be disrupted when I change providers (or
when my provider alters the prefix temporarily allocated to me).

  | I am similarly disturbed that there are people who want to
  | specify site-local addressing because they think it will offer
  | the provider-independence currently offered by NAT.

I have no idea who those people are.   I am surprised that you (seem to)
believe that NAT offers provider independence.   It doesn't.   What it
does is allows the renumbering task to be simplified (quite a lot usually
in that case) - but that's all.   Site locals do the same, they allow the
renumbering task to be simplified, as there's one less thing that needs
to be worried about.

By no means do they solve the entire problem, they're just one nail.
But they (or something essentially equivalent) are required, something
needs to provide the kind of address stability that anyone can now have
in IPv4 using NAT.

We could recommend NAT for IPv6, that would certainly do it, but I hope
we don't, there has to be better ways than that.

If you think you have a better way than site locals, please describe it.

I don't want to know what you think the problems with site locals are,
I know they're not perfect.   I want to know what is the better solution.
Until there is one, site locals need to remain.

kre