On Apr 4, 2012, at 12:39 AM, Greg Daley wrote:
Renumbering in small organizations _is_ a big deal, especially when they
don't have in-house skills to manage systems.
I tried to look at that in RFC 4192. The question isn't whether it's a pain or
not; it's what makes it painful. If building and deploying a protocol that will
automagically renumber devices in a network will solve that, let's spruce up
RFC 2894 and make it happen.
In the discussions we had, it turned out to be that any automated service
solved the easy 99% of the problem, whether it's done using DHCP, and fancy
protocol, or a database back end. The real issues in renumbering aren't so much
the process of distributing prefixes (IPv6 networks, route maps, ACLs, and so
on) as much as dealing with broken software that makes silly assumptions about
addresses - that an address that is meaningful to me is meaningful to you, that
addresses once assigned never change, that a host or service has exactly one
address, that given that knowledge one doesn't need to worry about names, and
so on. Unfortunately, when one actively goes out of one's way to bypass
anything that would allow us to automate the procedure, automated procedures
can't help us.
I agree that renumbering is a pain. I'm not certain what we could do in the
protocol that fixes "stupid".
That said, I'll put in a plug for RFC 6296, or for ILNP. ILNP would be a
superior solution if I thought I could change TCP and UDP; I think RFC 6296 is
a pragmatic step that is actually deployable (and deployed).