To:
dnsop@cafax.se
From:
Tim Chown <tjc@ecs.soton.ac.uk>
Date:
Wed, 30 Jul 2003 15:45:35 +0100
Content-Disposition:
inline
In-Reply-To:
<3F27D0AF.7010409@ehsco.com>
Mail-Followup-To:
dnsop@cafax.se
Sender:
owner-dnsop@cafax.se
User-Agent:
Mutt/1.4i
Subject:
Re: avoiding proxies
On Wed, Jul 30, 2003 at 09:05:35AM -0500, Eric A. Hall wrote: > > In the case of RA (as I understand it), the clients and servers couldn't > function without a proxy in any event, even if they were on the same > subnet. In the case of DHCP, a client and server could function in the > absence of a proxy if they were both on the same subnet, but not in other > scenarios. Stick the ~discovery capability into the service itself, and > this is all avoided. Wouldn't the RA always be local and on link though, as each link will have at least one router for off-link traffic? Thus there would not need to be an RA proxy as such. There would be a need for DHCP forwarding though, as one would be unlikely to have a DHCP server on every link, unless all routers included DHCPv6 Lite support. I think many enterprise networks now don't have clear L2 and L3 equipment boundaries. Functionality is combined, and to some extent blurred. In our enterprise network, which uses a lot of combined L2/L3 equipment, our core "switch" can forward DHCP requests from routed VLANs. It could equally handle RAs on links too, of course. I note that in many enterprises DHCP is not used. Hosts are manually configured, yet such enterprises have well known DNS resolvers. At our own enterprise (university) where we have DHCP we use it for the basic network config stuff (that is handled in v6 by RAs), plus DNS, DNS search path, NTP, NIS and some boot parameters. I don't know how typical that is in a campus department. The interesting question is how the DNS resolver (and/or other) information is configured to the router, for the RA method to work. Tim #---------------------------------------------------------------------- # To unsubscribe, send a message to <dnsop-request@cafax.se>.