On 3-jun-2005, at 0:00, Templin, Fred L wrote:
At the risk of covering old ground,
Yeah, we wouldn't want that. :-)
one question I had is whether a client must wait to receive an RA
before initiating stateless or stateful DHCPv6?
How would it know the value of the M and O bits if it didn't?
Asked another way, can DHCPv6 still be used if there are no
advertising routers on the link?
I suppose nothing is impossible, but it won't buy you much. If you
get an address, you can't use it, since DHCPv6 can't provide a
default gateway, like DHCPv4 does. Being able to learn additional
information is of questionable value, using multicast DNS / zeroconf
would be much more appropriate in an environment where there is no
global connectivity.
To an even more speculative question, if we had it all to do over
again would it be possible (or desireable) to design things such
that DHCPv6 and IPv6 ND messages could be "piggybacked" within a
single packet?
There is obvious synergy here.
I think we can still do this, and it could be very simple: define a
new RA/ND option that carries DHCPv6 options. Routers could inject
locally configured information, or do some form of stateless DHCPv6
and copy the results to RAs. (Stateful information must not leak out
through RAs, of course.)
Note that information in RAs is broadcast periodically so it takes up
some bandwidth regardless of whether it's needed. On the other hand,
hosts don't have to query for this information individually, which
saves bandwidth, especially on a shared medium such as 802.11.
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
