Re: [ipv6] Node Requirements: Elevating DHCPv6 from MAY to SHOULD

[Mohacsi Janos <[e-mail]>]

Hi,

You quoting is misleading - you getting out of context my answers.


On Tue, 31 May 2011, Markus Hanauska wrote:

>
> On 2011-05-31, at 11:57 , Mohacsi Janos wrote:
>
>> What about the ordering, if you get more than one DHCP addresses?
>
> How would this be any different to the situation as we have it today? 
> It's rather strange arguing to say something introduces a problem, if 
> this is not a new problem, but one that is in existence already by the 
> current standards. In whatever way this situation is handled today, it 
> would be handled the same way, I'd say.

Yes. Your address list has 4 element, mine list can have 5 or even more. 
You ordering is incomplete. I agree, It has to be handled in the same way:
bind() to proper address.

>
>> From operating system point of view all the addresses are the same.
>
> This is not quite true. If I have a SLAAC address with interface ID and 
> one with Priv Ext, all systems I know prefer the one with Priv Ext for 
> outgoing traffic, unless a client/server software explicitly picks 
> another one. The same thing could be the case for DHCP addresses.

This is not quite true either. You can select in some OS, to prefer 
privacy address or not. For DHCP, I don't see difference between static 
and DHCP assigned.

>
>>  The RFC 4941 states, that usage of Privacy enhanced addresses can be 
>> selected based on the destination addresses.
>
> Yes, but there is nothing special about this. Every address can be 
> picked according to the destination address simply by adding a routing 
> table entry for a certain destination or destination range; this also 
> works with IPv4.

Will it work this way on all the operating systems?

>
>> Not with SLAAC, but static addresses.
>
> My impression was the consent on this list is that static addresses are 
> evil(TM), since it is the most error prone way of assigning addresses on 
> a network. And I cannot really disagree, there is some truth in that 
> statement. Also it is more painful to remotely log into hosts and 
> manually assign them IPs, instead of managing all this in a central 
> place: The DHCP server config.

I am referring here the assigning static addresses to stable servers (to 
avoid problems of MAC address changes). For address assignment 
point of view static and DHCP is the almost the same. DHCP slightly worse, 
since you are relying operation of your DHCP server. Or slightly better, 
you have the assigned addresses in one database.

Regards,
 	Janos Mohacsi