On 26-mei-2005, at 22:24, da Silva, Ron wrote:
This seems way to complicated though, and it would be much more
deterministic for a CM to simply do #4 in all cases which would
make the
M/O bits useless.
You know what's really deterministic? Manual configuration.
There is a baby in this bath water...
The problem with having different bits indicate different
capabilities is that you get lots of permutations. So when a
significant number of those don't make sense, having individual bits
probably isn't the right choice.
To me, assuming the current specs, the following would make sense:
Configured for always-full-DHCPv6 on this interface? -> yes -> do
full DHCPv6
|
no
|
Timeout waiting for RAs? -> yes -> do full DHCPv6
|
no
|
M bit set or no prefixes with AAC set? -> yes -> do full DHCPv6
|
no
|
O bit set or always-stateless-DHCPv6 on this if? -> yes -> stateless
DHCPv6
A stateless DHCPv6-only client would probably just look at the O bit
and not do stateless DHCPv6 if just the M bit is set.
Anyone believe that, with current knowledge and experience,
implementing the above would be a bad idea?
Please realize that even if something isn't useful in _your_ use
case, that doesn't mean it's never useful. It's very important that
people can open up their laptops, PDAs, smartphones and whathever,
and it can talk to the local network _without_ _special_
_configuration_. I.e., I want to be able to bring my laptop to work
where DHCPv6 is the only game in town, check my mail using a public
hotspot over coffee with just stateless autoconfig and then go home
and connect to my broadband connection using whatever comes out of
the wall. And all this without having to change any settings.
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
