[dhcwg] Re: prefix length determination for DHCPv6

[Iljitsch van Beijnum <iljitsch at muada.com>]

On 14-aug-2007, at 12:38, tom.petch wrote:

> > Which got me thinking: wouldn't it be great if a single message from
> > a host could serve as both a router solicitation and a DHCP solicit
> > message?

> Yes, as long as the router 'server' and the DHCP server get their  
> act together
> and present a single coherent reply to the host:-)

If they tell the host the same thing then obviously we don't need two  
protocols.

I don't think there is necessarily a problem with different sources  
telling a host different things. But what we need is some rules about  
how the host is supposed to react to this situation.

Also note that today, this is already reality: you can have multiple  
routers. Hopefully, they all broadcast the same prefixes (if they  
don't, you could use a prefix from router A over router B and run  
into filters, unfortunately, for unexplained reasons, OS makers seem  
to think this is appropriate behavior) but obviously they'll be  
advertising different default gateway addresses. There are no rules  
about which a host should pick, but we do have dead neighbor  
detection which makes sure that only working ones are used.

Back to my original point: a mobile host isn't going to know whether  
a network it connects to is using stateless autoconfig, DHCPv6 or  
both. Sending discovery messages for both is probably wasteful in  
most situations, and sending discovery messages for only one will  
lead to delays if it chose the wrong one. So having one discovery  
message that can lead to replies of both kinds would be useful.  
Additionally, DHCPv6 servers could determine that a host doesn't need  
stateful information and the stateless information is in RAs, so it  
doesn't have to reply to certain request or requests from certain hosts.

_______________________________________________
dhcwg mailing list
dhcwg at ietf.org
https://www1.ietf.org/mailman/listinfo/dhcwg