[dns-operations] 答复: L-Root IPv6 Address renumbering

[Doug Barton]

On 03/14/2016 05:51 PM, Terry Manderson wrote:
>
> On 15/03/2016, 1:39 AM, "Robert Edmonds" <edmonds at mycre.ws> wrote:
>
>>
>> Hi, Terry:
>>
>> I don't think anyone asked this in the thread, but can you talk about
>> what the underlying reason is for the re-numbering? Given all the
>> operational burdens mentioned above, it sounds like something that
>> wouldn't be considered lightly.
>
>
> Hi Rob,
>
> Great question.
>
> It wasn't considered lightly, there has been considerable time thinking
> through this as to what real benefit we (and by "we" I then include all of
> the consumers of L-Root) get from it.
>
> The rationale behind the renumbering is that the current IPv6 prefix,
> 2001:500:3::/48, comes as a direct minimum assignment. We were unable to
> expand that allocation to a /47. What we (and I'm sure others) have
> noticed is that ISPs and transit providers filter on allocation
> boundaries, and constrain RIBs to the minimum allocation size. So our
> reach for doing traffic engineering with a /48 is quite limited, noting
> the first step in the BGP route selection state machine is generally the
> prefix length.
>
> The new address (2001:500:9f::42) sits in the prefix 2001:500:9E::/47,
> which then allows us to get 'reach' of the more specific 2001:500:9F::/48
> for traffic engineering purposes when we need it across the L-Root
> constellation on IPv6.
>
> We do this already on IPv4 using 199.7.82.0/23 and 199.7.83.0/24.
>
> Hope this helps.

Very useful, thanks Terry!

Doug