[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]


To: Bob Hinden <hinden@iprg.nokia.com>
CC: Tim Chown <tjc@ecs.soton.ac.uk>, dnsop@cafax.se
From: Doug Barton <DougB@dougbarton.net>
Date: Fri, 18 Jul 2003 09:42:27 -0700
Sender: owner-dnsop@cafax.se
User-Agent: Mozilla/5.0 (X11; U; FreeBSD i386; en-US; rv:1.0.2) Gecko/20030208 Netscape/7.02
Subject: Re: Stepping back on the DNS discovery discussion



Bob Hinden wrote:
> Tim,
> 
>> There was certainly a significant hum in this direction in the meeting.
>> Both DNS Lite and RA discovery methods can exist; each has an application
>> domain/scope.
> 
> 
> I agree with this view.  I think we should proceed with the RA approach 
> for learning the address(es) of the recursive DNS servers and optionally 
> a search list.

This is exactly the kind of feature creep that bothers me about the RA 
dns discovery approach. The proponents' argument boils down to, "Well we 
have 3 things, so it'll be easy to add a 4th (resolver[s])." The logical 
progression of that is that if you're going to add the 4th anyway, it's 
not really useful without the 5th (search list), and now I've got 
everything else I need, can't I please have the 6th (widget) too? 
Please? Pretty please?

Speaking as a client system implementor, I can almost see situations 
where the RA approach for stateless autoconfiguration of IP mask and 
gateway would be useful. However, by feature creeping RA you're 
basically forcing me to implement two autoconfiguration methods for 
every client, because if RA becomes "just robust enough" there are going 
to be networks where there is no dhcp.

Speaking as a resolving name server/dhcp configurator, I can easily 
imagine situations where the routers are pointed at one set of resolving 
name servers, and I want the clients pointed at a different set. This 
increases the complexity of the RA approach one notch already, and I'm 
just shooting from the hip here. I'm sure others can think of more 
complex problems.

The dhcp(lite) mechanism already exists to do everything that the RA DNS 
discovery could ever imagine doing, plus lots more. It's a scalable 
solution with a known implementation path. I fear that further muddying 
of the water is detrimental to both the long term goal of robust IPv6 
networking, and the short term goal of encouraging IPv6 adoption and 
deployment.

Doug

-- 
Angel:  We need you to distract the vampires.
Buffy:  Right.
Xander: What are you going to do?
Buffy:  I'm going to kill them all. (Walking away)
	That oughta distract them.
   "When She Was Bad" - Buffy the Vampire Slayer, Season Two Episode One


#----------------------------------------------------------------------
# To unsubscribe, send a message to <dnsop-request@cafax.se>.

Home | Date list | Subject list