need for DHCPv6 [Re: Geoff on IPv4 Exhaustion]

[Ralph Droms]

On Nov 17, 2011, at 9:47 AM 11/17/11, Doug Barton wrote:

> On 11/16/2011 15:53, Brian E Carpenter wrote:
>> On 2011-11-17 12:05, Doug Barton wrote:
>>> On 11/16/2011 14:56, Brian E Carpenter wrote:
>>>> On networks I know, including the one I used to
>>>> be responsible for so long ago that DHCP was a new feature, *exactly* the
>>>> same team controls the routers, the DHCP servers and the DNS servers.
>>> 
>>> I think part of the difference lies in the size of the enterprise in
>>> question, and part lies in the evolution of those roles in the "typical
>>> enterprise IT shop" over time. The operator community (including me) has
>>> been saying for over a decade that IPv6 needs DHCP parity with v4, in
>>> part for this exact reason.
>> 
>> Yes, and I suspect the DHC WG of the IETF would agree with you
> 
> Actions speak louder than words.

There are no constraints on actions brought to the dhc WG, which uses the same process as any other WG: Write a draft, bring it to the WG and ask for review/adoption.

If you're referring to options for carrying a default route and on-link prefix information, Thomas Narten and I wrote such a draft, brought it to the 6man and dhc WGs, and it did not generate enough interest to warrant moving it forward.  Details available if you're interested...

> 
>> (although 'parity' doesn't mean identical options, because the detailed needs
>> are different).
> 
> Right, that's why I have been using that word carefully.
> 
> I said 10 years ago that what we should do is simply copy over all the
> relevant options from DHCP for IPv4. I was told that it wasn't necessary
> because what would happen is that if an option was relevant, someone
> would write a draft and make it happen.

Overly simplistic.  Many DHCPv4 options are unnecessary (DHCPv4 option code 10 is the canonical example) or need to be modified for DHCPv6.  Rather than taking on unnecessary work, the dhc WG has added options to DHCPv6 as required (66 options defined so far).


> Oooooooookey dokey. Nice plan
> (?!?) but it didn't quite work out that way.

What's missing?

- Ralph

> 
>> The intention, as I recall it, of starting off in IPv6
>> by defining RA, ND and SLAAC was to allow plug-and-play for simple, and
>> possibly disconnected, networks.
> 
> Personally, I always thought that was a solution looking for a problem.
> However that plan also included, "RA for routing only, DHCP for
> everything else."
> 
>> What we keep arguing about is the exact scale of network above which
>> DHCPv6 is a necessity. I'm not even sure that's an argument worth
>> having.
> 
> But, it is. Because by not following the original plan (however
> misguided) we've gotten into the mess that mif is currently trying to
> deal with regarding reconciling conflicting DNS config information from
> DHCP and RA. That should never have been allowed to happen, and many of
> us tried to explain (in nauseating detail) why it was a bad idea.
> 
> The rational path has always been RA for routing only, and parity for
> DHCP (including network/gateway/etc., IOW, "DHCPv6 only"). Obviously the
> former is now unattainable, but the latter should be done ASAP.
> 
> 
> Doug
> 
> -- 
> 
> 		"We could put the whole Internet into a book."
> 		"Too practical."
> 
> 	Breadth of IT experience, and depth of knowledge in the DNS.
> 	Yours for the right price.  :)  http://SupersetSolutions.com/
>