On 17-aug-2007, at 0:51, Hemant Singh (shemant) wrote:
OK, Iljitsch. But hasn't Bernie already responded to this one? What if
prefix length is fat-fingered at the DHCPv6 server by the admin and a
length L1 is sent in DHCPv6 messages but the router RA sent a PIO for
the same prefix with prefix length L2. So as Bernie says, who wins?
What
does the host now do?
(Can't seem to remember what PIO stands for...)
What if someone misconfigures the dot zone file and the entire DNS
goes down?
Mistakes happen. There is no cure for stupidity. Etc, and so on, ad
nauseum.
The problem that I have is that if an address without a prefix length
becomes available, what do I do? Then I need to do different things
depending on what other information is available and redo all of this
when new information becomes available. It's a mess.
If a DHCPv6 server tells me I have address 2003::2003/68 and then I
get a router advertisement with a prefix option for 2003::/60 then I
know that both 2003::/68 and 2003::/60 are directly connected to my
interface. That the two happen to overlap is somewhat curious, but
otherwise no big deal.
I.e., in practice, the shortest prefix "wins".
Further, isn't sending prefix length in DHCPv6 messages duplicating
what
RA PIO prefix length field already is capable of doing?
Since when is there a requirement that a piece of information may not
be learned through two ways independently?
The point is that if a DHCPv6 server gives out addresses without a
prefix lenght, it only does half a job. Sure, you can let another
protocol pick up the slack and in theory that may not seem like a
problem, but the trouble is that in practice, different protocols run
and possibly fail at different times so the permutations increase
rapidly while if you just put it in the DHCPv6 message and get it
over with, life is simple and we can all go home early.
Lastly, what huge problem is sending prefix length in DHCPv6 messages
going to solve that is not solved by ND RA, RA PIO, and existing
DHCPv6?
Not a thing until someone is going to start USING DHCPv6 address
assignment, I'd say.
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
