Re: Architectural argument for RAs [Re: [69ATTENDEES] DHCP]

[Doug Barton <dougb@dougbarton.us>]

On Fri, 21 Sep 2007, Iljitsch van Beijnum wrote:

> On 21-sep-2007, at 1:26, Brian E Carpenter wrote:
>
> > > For example, learning default gateways via Router
> > >   Advertisements provides perfect fate sharing.
>
> I was saving this remark up for when we get to see the "do everything with 
> DHCPv6" draft, but this is as good a time as any:
>
> If we decide to go forward with putting gateway addresses in DHCPv6 to gain 
> administrative control over which gateway a host uses, in my opinion, the 
> only way we can do this without the potential for serious harm is to make 
> require that the same gateway address is both available in DHCPv6 and in RAs. 
> So the DHCP information is more like selection mechanism between gateways in 
> RAs than the authoritative source of the information itself.

Well at least you're finally making your argument explicit. You want RAs 
to be mandatory for everyone, but you don't want us to mandate the use of 
DHCP for you. Okey dokey.

> And I agree with James' vision: having servers announce their service through 
> a discovery mechanism is more robust than having an outside system enumerate 
> them.

This part I agree with. I think mDNS has a very bright future, and makes 
total sense in the "stateless autoconfiguration" case. "Well known 
addresses" is a path we should not go down.

Doug


--

	If you're never wrong, you're not trying hard enough