IPv6 Pain Experiment

Doug Barton dougb at dougbarton.us
Fri Oct 4 17:10:27 UTC 2019


On 10/4/19 7:45 AM, Warren Kumari wrote:
> On Fri, Oct 4, 2019 at 5:13 AM Masataka Ohta
> <mohta at necom830.hpcl.titech.ac.jp> wrote:
>>
>> Doug Barton wrote:
>>
>>> And even
>>> if you do need to change providers, once you have your addressing plan
>>> in place all you have to change is the prefix.
>>
> 
> This is the same as saying "If you need to change providers in IPv4,
> once you have your addressing plan in place all you have to do is
> change the prefix", or "To build the Eiffel Tower, all you have to do
> is bolt bits of metal together" -- it's technically correct*, but
> handwaves away the actual complexity and scale of work.
> Yes, you (clearly) can renumber v6 networks, and it's *probably*
> easier than renumbering v4, but "just change the prefix" oversells it.

I was assuming that this audience understands the relative complexity of 
changing the network parts of the address, and leaving the subnet and 
host parts in place.

And by and large, it is not true that you can do this with IPv4. You 
might occasionally get lucky with it, but that would be the exception, 
not the rule.

As for your Eiffel Tower example, the design and architecture are the 
pieces that would already be in place as part of your networking plan, 
so in a sense we're talking about RE-building the Eiffel Tower by taking 
off one bit of metal (the old network) and bolting another piece in its 
place. Not building it all over again from scratch.

So you can credibly accuse me of underselling the renumbering effort, 
but don't engage in hyperbole to try to balance the scales.

Renumbering has its pain points regardless of the protocol, but if 
you've designed your network correctly IPv6 renumbering is considerably 
easier than it is in IPv4.

Doug




More information about the NANOG mailing list