[arin-ppml] ARIN Draft Policy 2011-5: Shared Transition Space

[Owen DeLong]

On Jul 5, 2011, at 2:40 PM, William Herrin wrote:

> On Tue, Jul 5, 2011 at 5:30 PM, Owen DeLong <owen at delong.com> wrote:
>> On Jul 5, 2011, at 2:24 PM, William Herrin wrote:
>>> On Tue, Jul 5, 2011 at 5:13 PM, Owen DeLong <owen at delong.com> wrote:
>>>> On Jul 5, 2011, at 2:01 PM, William Herrin wrote:
>>>>> 2. Router interface numbering is a potential use case. Filtering of
>>>>> RFC1918 is too widespread to overcome, so using it outside the NAT
>>>>> breaks path MTU detection. That's not inherently true of this new
>>>>> space and nonfiltering could be encouraged in a way that renders it
>>>>> usable in a few years.
>>>> 
>>>> Not really... How would you know which ISP to route the packet
>>>> back to?
>>> 
>>> You don't care. The packets you care about receiving from the router
>>> are ICMP type 3 (destination unreachable) and ICMP type 11 (time
>>> exceeded) . Those flow only one direction -- from the router back
>>> towards you.
>> 
>> If a packet is inbound towards me and needs fragmentation, for PMTU
>> discovery, then, an ICMP type 3 would be sent back towards the
>> originator of said packet.
> 
> Correct, but the originator doesn't care what address that packet
> comes from and doesn't attempt to reply to it.
> 

I tend to doubt anyone would be any more likely to accept these addresses
across their border as source addresses than RFC-1918.

Owen