On 06/11/2012 08:15, Fernando Gont wrote:
> On 11/05/2012 09:05 PM, Brian E Carpenter wrote:
>>> * It looks like the ultimate goal of these tokenized addresses is that
>>> of renumbering? Am I right? This is pointed out in the abstract, but
>>> doesn't seem to be so clear form the Intro.
>> As Fred Baker has pointed out in the 6renum context, there is only a small
>> difference between "renumbering" and "numbering", and every network gets
>> numbered at least once in its life, and more often if powered off and on
>> again.
>
> Huh? :-)
>
> If you generate your addresses according to traditional SLAAC, and the
> advertised autoconf pprefix is always the same, where's the renumbering?
At least for a homenet or small office buying basic service,
there's no guarantee you will get the same prefix back if you
power cycle your CPE.
>>> * When it comes to servers, I guess they'd nevertheless need to update
>>> the corresponding DNS entries... So this proposal seems to be more
>>> targetted at, e.g., ACLs?
>> If dynamic DNS update is in use, there would be no manual action needed.
>
> The point is, if you're updating the DNS entiries anyway, why would you
> care about the special magic (i.e., tokenized IIDS)? -- the meat seems
> to the in the renumbering case.
Site managers of enterprise networks like predictability, and
they don't want things to change in the case of a swap of the
physical server, and of course they don't want to manually
configure an IPv6 address on the new box.
However, I agree: we need renumbering to become a routine
operation, so that is a main driver for parameterising address
config as much as possible.
>> Indeed, a configuration system could know that <prefix-N><token-M> is
>> the address for a given server, and everything including the AAAA record
>> could be parameterised on that basis.
>
> I'd set the DNS aside -- at least this I-D doesn't mention any mechanism
> to notify the DNS of the updated *prefix* (and you probably wouldn't
> want that, anyway).
It's a separate problem, but yes, I would want it automated (and
secure, of course). We should aim at typing in hexadecimal
numbers as little as possible. Tokenised IDs are a tool for
that, and so are parameterised prefixes.
Brian
--------------------------------------------------------------------
IETF IPv6 working group mailing list
[email protected]
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
