[ipv6hackers] "Stick to limited IPv6 deployments, businesses warned"

Owen DeLong owend at he.net
Mon Sep 10 09:08:05 CEST 2012


On Sep 9, 2012, at 16:35 , Fernando Gont <fgont at si6networks.com> wrote:

> On 09/08/2012 04:49 PM, Owen DeLong wrote:
> [....]
>> I say implement it all and let each environment pick the solution that works
>> best for them.
> 
> FWIW, I agree with your pov
> 
> BTW, my read of the religious wars in this area is that:
> 
> * Some guys say do not include more info in SLAAC -- if you need
> additional info such as DNS, you should be doing DHCPv6.

Yes, but they should be ignored, though other than RDNSS, I wouldn't
really want to add anything else and agree that if you need more than
the absolute basics, DHCPv6>SLAAC. The difference being that I
consider RDNSS an absolute basic and some of the zealots in this
argument do not.

> 
> * Other guys (anti-DHCPv6?) say "DHCPv6 should not be able to convey all
> the necessary information.. e.g., routs should be handled with RAs
> rather than DHCPv6".
> 

Yes, but they should also be ignored.

> The result of both factions is that, at the end of the day, you need to
> support both protocols even for simple auto-configuration stuff.
> 

The most unfortunate part of it is that anybody actually listened to either
faction.

> My take is that both autconf mechanisms should be able to handle at
> least basic configuration stuff (addresses, DNS, and routes), and then
> it should be up to the admin which one to use.
> 

Agreed.

SLAAC+RDNSS = perfectly functional for 99% of environments, actually.
DHCPv6 = more comfortable for people used to DHCPv4 and more useful in areas where more extensive autoconf information or control
	is required.

Simple as that, really.

I've used both protocols and feel that they both have places where they excel and places where they are less desirable.

Owen




More information about the Ipv6hackers mailing list