[arin-ppml] [Fwd: Draft Policy 2011-5: Shared Transition Space for IPv4 Address Extension]

[Joe Maimon]

Owen DeLong wrote:
> On Feb 25, 2011, at 10:02 AM, Joe Maimon wrote:
>
>    
>>
>> Owen DeLong wrote:
>>      
>>> On Feb 25, 2011, at 8:52 AM, Joe Maimon wrote:
>>>
>>>
>>>        
>>>>          
>>> You're requiring changes to both the CPE and the provisioning system that are impractical in the circumstance.
>>>
>>> Owen
>>>
>>>
>>>
>>>        
>> The only changes required are on the DHCP server. DHCP clients will continue to request leases all on their own. End users will continue to reboot CPE's all on their own.
>>
>> And this is only one option of many, a number of which can be employed in tandem.
>>
>> Joe
>>      
> You've made the assertion that the DHCP client will reject a lease for an address that it cannot install in
> its routing table due to conflict with the LAN segment. I am arguing that would be a change to the
> behavior of most CPE.
>
>    

I assume you have anecdotal evidence, just like I have anecdotal 
evidence of an ISP whose CPE defaults to 10.0.0.100/16, knocking out the 
DHCP client on the Corporate IOS with internal lan 10.4.x.x/27. Fix the 
mask, problem solved, remotely.

A solution doesnt have to be perfect, it simply needs to lower the 
problem ratio to manageable levels. Which this would. Which the others 
solutions would as well. Which combined they would do so even more.

> Yes, you can make the argument that CPE that doesn't behave that way is broken. Guess what,
> most CPE is arguably broken in at least one fundamental way. That's what happens when you
> buy routers for $30.
>
> Owen
>
>
>    
These are exactly the CPE's that are rebooted repeatedly by their users 
hoping to get them to work again.

More problem space covered.

Joe