DHCPv6-4.1.mumble

[bmanning at vacation.karoshi.com]

i did mention merging as an issue -IF- it was not possible to 
pass DNS server information.  to borrow from the DNS side of the
house, it is perfectly reasonable to return, over IPv4 transport,
data about IPv6.
the dhcp "Domain Nameserver" option should be able to return a series
of IP addresses that a client might use.  this capability is seriously
crippled by the presumption that the IPv4 version of DHCP can -ONLY-
return IPv4 addresses for this option and the IPv6 version of DHCP can
-ONLY- return IPv6 addresses for its functional equivalent of the
Domain Nameserver option.  The presumption seems false.

If, as asserted elsewhere in this thread, the IP address string to be
returned is unspecified then it should be fairly easy to assert that
any valid IP address string/structure should be valid. I am working on
patching the ISC code to do just this.

If this works, it will avoid the issues of data merging... for now.

--bill


On Thu, Jun 12, 2008 at 06:10:11AM -0400, Ralph Droms wrote:
> We've had variants of this discussion in the dhc WG.  The problem was  
> mentioned - I think by Bill - in this thread: how does the client deal  
> with information from multiple sources?  In this case, does the client  
> merge the list of DNS servers from the two DHCP sources into the local  
> DNS server list (e.g., resolv.conf).  What about manually configured  
> information - is it overridden or retained?  How is the information  
> updated; suppose the same server is both manually configured and  
> learned through DHCP, and then that server drops from the DHCP list?
> 
> The WG has concluded these issues are really best left to the client  
> implementation, as there is no way to describe and dictate a global  
> policy around this problem.
> 
> BTW, the discussion might be a good one to take up on the dhc WG  
> mailing list...
> 
> - Ralph
> 
> On Jun 12, 2008, at Jun 12, 2008,5:32 AM, Simon Hobson wrote:
> 
> >If I might throw in my 2d worth ...
> >
> >Getting back to the original problem, surely the cure is to fix the  
> >IPv6 allocation mechanism to properly configure the IPv6 side of the  
> >client ?
> >
> >In the general case I can't see the point of trying to make the IPv4  
> >client configure IPv6 stuff since there are no guarantees about if/ 
> >how the IPv6 side is configured - similarly for IPv6 clients  
> >configuring IPv4 stuff.
> >
> >Given that getting existing IPv4 clients to handle IPv6 stuff will  
> >require a LOT of clients to be modified, one could also question the  
> >wisdom of going down that road which is likely to take many years to  
> >get a spec agreed and then get vendors to support it.
> >
> >And of course, a DNS server running on IPv4 can offer IPv6 addresses  
> >etc.
> >
>