Le 01/08/2012 18:48, Karl Auer a écrit :
On Wed, 2012-08-01 at 17:58 -0700, Alexandru Petrescu wrote:
The choice today is more than b/w 'if M then DHCP is available
otherwise SLAAC'.
That has never been the choice. It's never been either/or, one
excluding the other.
Ok fair enough. The eisting M/O are like hints - suggestions.
Rock solid mechanisms exist to stop hosts doing SLAAC - the autoconf
flag, which DOES impose mandatory behaviour on the host, or,
worst-case, use of a non/64. And a rock solid method exists to stop
hosts doing DHCPv6 - just don't provide service to the subnet.
Right...
Neither mechanism is onerous, neither method is complex.
Both DHCP and SLAAC advanced towards doing what the other does,
and it is _still_ possible to need some features from one and some
from the other - it's still impossible to use just one for
everything. (eg RA doesnt PD, DHCP doesnt MTU, and more)
You are talking here about what information is delivered via the two
mechanisms, not the operations of the mechanisms themselves.
Right. Let me try to better explain.
In a static environment, where the PC always wakes up on the same desk,
this avialability of DHCP vs SLAAC is easily decided once for several
months. This could happen easily with just M/O, or simply by the
absence of DHCP Server.
But in mobile environment, a mobile router moving aroung attaching to
other mobile router needs to be able to face quickly various ways of
delivering this configuration data.
If one considers simply the triplet address-defroute-prefixdelegated
then there are about 8 possibilities to deliver that to a Router, by
using DHCP and ND combinations. Which of the 8 should the router try first?
One neighboring vehicle would deliver it
address-defroute-prefixdelegated by DHCP, another one by SLAAC and yet
another one by a combination of the two.
In order to test for the availability of which of DHCP/ND delivers which
of address-defroute-prefixdelegated, one would send several RSs and
several DHCP Requests. This may be too much for routers which
constantly move. Ideally, such a Router would just send an RS, receive
some capability, and subsequently executing DHCPReq depending whats in
that RA.
This would require more than just M/O flags.
And yes I agree this would be more that what a wide agreement could be
reached on M/O...
Just some thought.
Alex
Regards, K.
--------------------------------------------------------------------
IETF IPv6 working group mailing list
[email protected]
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
