Re: [dhcwg] [v6ops] Fwd: New Version Notification for draft-link-dhc-v6only-01.txt

[Roy Marples <[e-mail]>]

On 12/12/2019 23:09, Bernie Volz (volz) wrote:
> That is a good find. I didn’t recall this.

I concurr, a veru good find.

> 
> On the surface, it looks like this document does a similar thing to the 
> v6only option (stop DHCP). And, if yiaddr was 0, this would require 
> checking both the autoconfigure and v6only to see how to proceed?

I still don't understand why we need to check anything if yiaddr is 
unspecified.

> 
>     DHCPOFFERs with a 'yiaddr' of 0x00000000 will only be sent by DHCP
> 
>     servers supporting the Auto-Configure option when the DHCPDISCOVER
> 
>     contained the Auto-Configure option.  Since the DHCPDISCOVER will
> 
>     only contain the Auto-Configure option when a DHCP client knows how
> 
>     to handle it, there will be no inter-operability problems.
> 
> It is likely something that the authors of draft-link-dhc-v6only will 
> need to think about if a yiaddr of 0 is allowed for that case? 
> Personally, another good argument not to use a yiaddr of 0 as it has a 
> special semantic per RFC2563.

Really?
Are you entirely sure that your middleware box is going to allow the 
unspecified address if this option is present AND allows it?

Sorry but that is just going to far.
RFC's have been updated in the past to allow for more conditions on 
pre-existing behaviour - this is no different.

I have no idea what the bee in your bonet is about forcing an address at 
this point as clearly we have RFC's which allow the unspecified address 
in yiaddr. There is clear vocal opposition to the proposal to force an 
address where there is no need to be one.

 > Thanks much for pointing out this document!! (Not sure how widely
 > implemented this is as no one else mentioned it.)

dhcpcd-6.3.0 supported this in 2014.
I would imagine the support for it is quite wide.

Roy