Re: Distribution of RFC 3484 address selection policies

[Greg Daley <[e-mail]>]

Hi Fred.

Fred Baker wrote:
> On Aug 8, 2005, at 7:24 PM, Greg Daley wrote:
> 
>> I'm not sure anyone is doing it, but renumbering is applicable  there 
>> as a means of providing information about which prefixes are  valid.
> 
> 
> One of the outcomes of the v6ops WG last week was the observation  that 
> the Router Renumbering Protocol is not widely implemented, and  when 
> renumbering-in-anger persuant to
> 
>   http://www.ietf.org/internet-drafts/draft-ietf-v6ops-renumbering- 
> procedure-05.txt
>   "Procedures for Renumbering an IPv6 Network without a Flag Day",  Fred 
> Baker,
>   18-Mar-05, <draft-ietf-v6ops-renumbering-procedure-05.txt>
> 
> was tested by the 6net folks, it was found wanting in a variety of  
> ways. Basically, it made the special case where you want to  distribute 
> a new aggregated prefix and maintain the same subnet  number 
> distribution (such as "I have changed upstream provider")  pretty 
> simple, but didn't cover any of the other cases in which one  might want 
> to renumber, and didn't handle any of the nuances like  making sure this 
> also happened in route maps, aggregation filters,  qos classification 
> filters, etc.
> 
> As the discussion proceeded, we basically decided that renumbering  (and 
> more generally number distribution and number use policy) was  the 
> province of the network management folks, perhaps netconf. They  need to 
> decide how to manage a network, and if a protocol like RRP is  part of 
> that, specify a set of requirements for it. RRP at minimum  needs work, 
> and I would suggest it be declared "historic" and  replaced with a new 
> protocol if such a definition happens.

Sure.   I guess that if automated generation of policy mappings for
source address selection are required then it would be useful to
work on this in concert with the main thrust of network management.

Moving RRP to Historic may leave questions unanswered in peoples'
heads regarding the ISP change scenario, but I really think moving the
focus to SHIM6 (which provides host-to-host survivability of
renumbering events) may actually be the better course (Shim6 people can
take aim now).

What's reasonable here with regard to address usage is to provide
indications of the network's perception of differences between addresses
to hosts.   The actual task of managing the network's perception
isn't really any different in DHCPv6 or Router Discovery (it's either
automated or manual).  The hosts receiving the default policy aren't
ever aware of the mechanism used.

As you mention, RRP may not be a valid way to do this configuration
today (or in future), but config files, SNMP and misuses of RADIUS
(not actually recommended) may be.

Greg

--------------------------------------------------------------------
IETF IPv6 working group mailing list
[e-mail]
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------