[arin-ppml] IPv4 SWIP requirements (?)

[hostmaster at uneedus.com]

I have been using v6 since 2007, and everything that was ever stated in 
the RFCs and in practice always recommended that assignments align on a 
nibble boundary.  Having had many v4 assignments less than /24, I know of 
the CNAME tricks used.  I never had a non nibble aligned v6 assignment, as 
I always received a /48.

Having never used anything outside of a nibble boundary, I never actually 
considered v6 NS records outside of a nibble boundary are much easier than 
the v4 CNAME trick.  It appears that no more than 8 NS records would be 
required for any chosen boundary.

Based on comments so far, most agree that a /48 should be SWIP'ed since it 
is routable on the internet, and since so far the majority seems to think 
that /56 is small enough to not require SWIP, this leaves 7 choices of /49 
to /55 to set the limit for SWIP in the Draft.

I still think a nibble boundary aligned value of "more than /52" or "more 
than a /56" will be the best choices in that range to place in the draft, 
rather than a non nibble aligned value.

Albert Erdmann
Network Administrator
Paradise On Line Inc.

On Thu, 15 Jun 2017, Owen DeLong wrote:

>
>> On May 25, 2017, at 21:02 , hostmaster at uneedus.com wrote:
>>
>> This proposal was intended to try to bring the v4 and v6 world together 
>> on the same policy.  Because of the nibble boundary rule and rDNS, on 
>> the v6 side, there are really only 5 choices in network size: /48, /52, 
>> /56, /60 and /64 without having to do non-standard CNAME tricks used 
>> when subdividing the rDNS for more than one customer in a /24 of v4.
>
> Actually, this isn’t true.
>
> It just means _EXTRA_ NS entries.
>
> For example, 2001:db8:beef::/48 can be delegated as f.e.e.b.8.b.d.0.1.0.0.2.ip6.arpa. as a single NS record set.
>
> However, to delegate 2001:db8:beef:e000::/49, one needs two NS delegations:
> 	e.f.e.e.b.8.b.d.0.1.0.0.2.ip6.arpa.
> and	f.f.e.e.b.8.d.b.0.1.0.0.2.ip6.arpa.
>
> No CNAME requirement at all.
>
> CNAME tricks would only come into play if delegating reverse for prefixes longer than /124. (Just as in IPv4, they only come into play longer than /24).
>
> Owen
>
>