IPv6 Pain Experiment

Owen DeLong owen at delong.com
Sat Oct 5 03:35:18 UTC 2019



> On Oct 4, 2019, at 20:23 , Owen DeLong <owen at delong.com> wrote:
> 
> 
> 
>> On Oct 4, 2019, at 16:48 , Michel Py <michel.py at tsisemi.com> wrote:
>> 
>>> Owen DeLong wrote :
>>> How would you have made it possible for a host that only understands 32-bit addresses to exchange traffic with a host that only has a 128-bit address?
>> 
>> With some kind of NAT mechanism, naturally.
>> Which is not possible with the current IPv6 address format, if you want something stateless and that does not rely on DNS.
> 
> Well, what address format would you propose that would make it better? Let’s talk actual workable detailed proposals rather than just hand-waving.
> 
> We already have a number of such solutions:
> 	NAT64
> 	464XLAT
> 	B4/AFTR
> 	etc.
> 
>>> How would you have made a 128-bit address more human-readable? Does it really matter?
>> 
>> I have found it difficult to talk hex with people from other countries.

Sorry, hit send too soon. Won’t rehash previously posted content, but here’s what got missed…

In addition, hex makes it MUCH easier to do subnetting. Each digit aligns with a nibble boundary, so
instead of having to memorize/calculate 8 different powers of two ranging from 1-128, you only need
to memorize 4 ranging from 0-8. Further, given the bountiful amount of IPv6 space available, you shouldn’t
really need to subnet off nibble boundaries unless you really want to.

How many people do you know (as a percentage) that divide RFC-1918 space into non-octet-aligned subnets?

Remember the handy subnet calculators for IPv4 that broke down all the net mask possibilities:

/ 9,  /17, /25 — .0/.128 (0-127 and 128-255)
/10, /18, /26 — .0/.64/.128/.192 (0-63, 64-127, 128-191, 192-255)
…
/15, /23, /31 — .0/.2/.4/.6/.8/.10/.12/.14/.16/…

Yeah, compare that to:

/n % 4 =
0	— Aligns with nibble boundary.
1	— 0/8 (0-7, 8-f)
2	— 0/4/8/c (0-3, 4-7, 8-b, c-f)
3	— 0/2/4/6/8/a/c/e (0-1, 2-3, 4-5, 6-7, 8-9, a-b, c-d, e-f)

Subnetting is MUCH MUCH MUCH simpler in IPv6, especially if you follow the intended architecture/recommendations.

Owen




More information about the NANOG mailing list