IPv6 Pain Experiment

[owen at delong.com (Owen DeLong)]

> 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? Letâ??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. Wonâ??t rehash previously posted content, but hereâ??s what got missedâ?¦

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 shouldnâ??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 â?? .0/.128 (0-127 and 128-255)
/10, /18, /26 â?? .0/.64/.128/.192 (0-63, 64-127, 128-191, 192-255)
â?¦
/15, /23, /31 â?? .0/.2/.4/.6/.8/.10/.12/.14/.16/â?¦

Yeah, compare that to:

/n % 4 =
0	â?? Aligns with nibble boundary.
1	â?? 0/8 (0-7, 8-f)
2	â?? 0/4/8/c (0-3, 4-7, 8-b, c-f)
3	â?? 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