On Wed, 25 Jul 2012, Jethro R Binks wrote:
> Putting it another way; while I might chose to take the opportunity to
> do IPv6 numbering in a more logical structured way according to
> topologies, potential for route aggregation, and so on, I'm extremely
> unlikely to start changing the extant VLAN ids to match that neat IPv6
> addressing in many cases.
We have a whole pile of historical "cleverness" (and I'm a big one for it
myself, to help me identify the location of a VLAN or IP address, or guess
the IP of a router or link). Some of this survives the test of time and
others turn out to be massively wrong.
When we reinvented IPv6, going from a /48 to a /44, I decided I'd just
allocate the /56s to institutions (=> departments and colleges) in
sequential order. What might start out as fantastically clever at the
beginning (like dividing the routing table to summarise it) quickly falls
apart with the university moving departments out to west Cambridge, or
colleges requesting their primary connection moves to their new
out-of-centre site.
What we do (that I've sent round before) is here:
http://www.ucs.cam.ac.uk/network/ip/ipv6-at-cambridge/#university
... the large size of allocation for each institution is connected with
the highly decentralised nature of the University, so probably wouldn't
apply to others.
Personally, I regard IP addressing at a layer most people shouldn't care
about, as DNS sits on top and handles translating those nice names that
the marketing people like into those horrible numbers I force on the
system administrators in institutions.
As such, I pick IP addresses that fit how I want to organise things -
which means how I set up filters in OSPF/BGP import/export lists, or
access lists for management access.
We did have a discussion, when reinventing things, if we should try to
solve some of the problems we hear from people such as the University
Library - like putting various departments in different blocks of
addresses, so they could exclude them from journal lists. However, there
are many ways to slice the cake and I think it's essentially futile - it's
probably best not to try and be too accommodating because people
misunderstand what they're getting.
- Bob
--
Bob Franklin <[log in to unmask]> +44 1223 748479
Network Division, University of Cambridge Computing Service
|