It seems like there might be several incorrect assumptions here leading to over
thinking the issue.
1. Over a long period of time, will the size or number of subnets be
significantly different than today. Even today a bunch of our assumptions on
why subnets are created the way they are might be obsolete or close to it. For
example, broadcast domain size becomes less of an issue as broadcasts are used
less (presumably by better targeted multicast and smarter protocols) and device
performance increases. Maybe networks get smart enough to reorganize
themselves and their subnets to optimize performance or maybe there will not be
the need to do so.
2. Am I wrong or are we making some bad assumptions about
"routability" of subnets? It seems like we might be too concerned about the
minimum routable sized subnet when in reality that is just a policy decision
open to change over time. Also, the routing I do inside my home does not need
to be routable to the Internet at large. Summarization can occur in my home
and at the service provider level.
3. If a couple of users required an inordinate amount of subnets, why
not assign them an additional netblock? In the V4 world Fortune 500 global
networks are not treated the same as your grandma's cell phone. It's nice to
have the comfort of doing it with V6 but no real compelling reason not to have
more than one type of customer assignment. Why not size things for the usual
use case and add another allocation as needed? Seems really wasteful to do
anything else. The NICs don't give a global carriers the same assignments they
give a simple dual homed user why should you think all of your sub-assignments
would be identical?
4. This is not the last time in history that you might get to
reorganize your address space so don't try to plan for crazy edge cases. Given
the widespread use of DHCP (or any other automatic addressing technology) in
the V6 world it is archaic to think that an address change will be a big
inconvenience. If my device receives an address and auto updates DNS records
then I really do not care what that address is. Feel free to reorganize the
network whenever you want.
5. The car example does not work because I am assuming you don't drive
in your home. You probably roam all around the area and will be getting
dynamic assignments from whomever the provider is. Do you statically address
your iPhone when it is on the 3G/4G/LTE network?
Steven Naslund
Chicago IL
>>>Subject: Re: Dual stack IPv6 for IPv4 depletion
>>>On 9 July 2015 at 11:42, Matthew Huff <mh...@ox.com<mailto:mh...@ox.com>>
>>>wrote:
>>>What am I missing? Is it just the splitting on the sextet boundary that is
>>>an issue, or do people think people really need 64k subnets per household?
>>>One thing you're missing is that some of these new-fangled uses for IP
>>>networking will want to do their own subnetting. It's not "here's a subnet
>>>for the car", it's "here's a /56 for the car to break into smaller pieces as
>>>>>>required".
>>A /56 isn't 256 subnets, it's 8 levels of subnetting (or 2 levels, if you're
>>human and want to subnet at nibble boundaries). A /48 is 16 (or 4) levels. I
>>have four vehicles, so I'd want to carve out a /52 for "the car network" to
>>>>make the routing and security easier to manage, and leave room for
>>expansion (or for my guests...)
>>One more consideration for you: we're currently allocating all IPv6 addresses
>>out of 2000::/3. That's 1/8th of the space available. If we discover we've
>>messed up with this sparse address allocation idea, we have 7/8ths of the
>>>>remaining space left to do something different.
