Re: [69ATTENDEES] DHCP

[Shane Kerr <shane@time-travellers.org>]

Gert,

I should probably let this thread die, but I wanted to talk about this
point...

On Thu, Sep 13, 2007 at 08:37:29AM +0200, Gert Doering wrote:
> 
> Well, I fully agree with you.  "Networks differ, so do
> requirements".

I agree with this. But I disagree with the following characterization:

> I have some issues with some folks from the DHCP camp refusing to
> understand that there are operators that just follow a different
> operational model.

I wasn't around in "the good old days when IPv6 was being designed",
but my understanding is that IPv6 designers wanted to add
"autoconfiguration" to the list of improvements over IPv4. (This is a
bit silly, because everything other than "IPv6 has more addresses" has
been added to IPv4 - of course!) Because of this, they resisted
DHCPv6, and they continue to oppose DHCPv6.

I suppose I am "in the DHCP camp", but my take on it is not to oppose
network designs without DHCP, but rather that we should not attempt to
duplicate the functionality of DHCP in other protocols. So, you may
think "we can allocate addresses, all we have left is DNS
information!", but....

DHCP IS NOT JUST USED TO ASSIGN ADDRESSES!! It is a generic protocol
that hosts can use to find out information about the network that they
sit in. It also allows the DHCP server can update network
configuration (like reverse DNS, or firewall configuration, or routing
tables) based on host preferences.

So it may seem like DHCP folks are trying to impose one model of
networks. But the real goal is to avoid creating a second,
less-functional way of doing the same thing as DHCP already does in
the network.

We already have DNS information in router advertisements(*), so I
don't think it needs to be discussed any more. But expect to get
push-back from DHCP folks for whatever the *next* thing people try to
put into RAs is. Not because we are fighting against your network
design, but rather to avoid creating duplicate, complicated,
non-interacting protocols. We want to put the right functionality in
the right place.

(*) Okay, maybe only an experimental RFC:

http://tools.ietf.org/wg/dnsop/draft-jeong-dnsop-ipv6-dns-discovery-12.txt
http://www1.ietf.org/mail-archive/web/ietf-announce/current/msg03819.html

--
Shane