-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
james woodyatt wrote:
> On Aug 13, 2007, at 09:10, David W. Hankins wrote:
>> On Sat, Aug 11, 2007 at 05:51:09PM +0900, JINMEI Tatuya / 神明達哉 wrote:
>>>
>>> In any event, I hear that some DHCPv6 guys are planning to make a new
>>> draft that covers this topic. So I think it's better to hold off for
>>> now and wait for the document, rather than continue this thread
>>> with keeping possible misunderstanding or confusing about the base
>>> protocol principles.
>>
>> I agree with everything Jinmei said, but this in particular. Any
>> conclusions from these threads have a tendency to get forgotten -
>> it's better that we argue the position of a draft that an editor
>> can follow consensus.
>
> I agree with M. Jinmei as far as the quoted excerpt above goes.
>
> As much as I dislike DHCPv6 on general principle, I don't expect that
> IETF will abandon it in favor of a system that more coherently addresses
> the various separate concerns that DHCPv6 attempts to meet. I hope the
> DHC WG will specify requirements for routers sending advertisements with
> the M bit set that prevent any ambiguity from arising at the node during
> link configuration.
>
> p.s. I really wish IETF had deprecated DHCPv6 when it had the chance.
> Every last thing DHCPv6 does belongs in one of two other, separate
> domains, i.e. either in the sub-IP layer, e.g. EAP, or at the
> sub-application layer, e.g. DNS-SD. Instead, IETF chose a short-sighted
> and half-baked solution to a problem that really didn't need a rush job
> to solve, and the engineering community today is poorer for it.
We need DHCP.
DHCP is used by administrators to do a lot more than simply assign IP
addresses. It is an extremely flexible protocol, and works for lots of
different models of network topology and administrative setup. Any
protocol which did the work that DHCP does would end up looking
suspiciously like DHCP at the end (RFC 3315 gets my vote as one of the
clearest, easiest to implement RFCs, BTW).
I know there's a lot of anti-DHCP sentiment, especially in the IPv6
crowd. I think it is misplaced; DHCP is at worst a necessary evil, and
at best an elegant solution to the problem space where it sits.
Router advertisements cannot do what DHCP does. An RA can carry a
small bit of information, but as a full replacement, the idea falls
flat. You don't always want to broadcast the same information to all
clients. Some hosts want vendor-specific information that will never
be standardized into an RA. A few people may even want to authenticate
their hosts' configurations. And on and on...
At the most basic level, a computer connecting to the network expects to
get an IP address, routing information, and DNS information. Until *last
year*, there was no standard way to get DNS information from an RA. Now
that this "problem" has been solved (*), what next? What about other
basic things like updating the reverse DNS information? What about being
able to log when a given IP was in use for legal, or (far more
importantly) billing purposes?
(*) I put "problem" in quotes because you could use DHCPv6 for this
since 2003 or so.
It may be possible to force much of this information in to RAs, having
multiple protocols to do one thing is a bad idea, right?
We need DHCP - in IPv6 as much as in IPv4 - so can we please let this
idea that we can do everything in RA die? Please?
- --
Shane
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org
iD8DBQFGytuOMsfZxBO4kbQRAj1gAKCF2shHEhrbOU+4g8E53g1OhlHTMQCfQVfa
dgyOZkLI+EVd86rkYwmOTmM=
=DtH/
-----END PGP SIGNATURE-----
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
