Re: [OPSAWG] [v6ops] Heads up - Plea for allocating a /8 to ISPs

Hi Roger,

On 11Nov2010, at 17.31, Roger Jørgensen wrote:

> On Thu, Nov 11, 2010 at 3:30 AM, Rémi Després <[email protected]> wrote:
>> Hi, Ron,
>> 
>> Le 9 nov. 2010 à 13:26, Ronald Bonica a écrit :
>>> I am having a strange sense of déjà vu. About two years ago, we saw another
>>> proposal to allocate a /8 for private use to support another transition mechanism.
>>> We concluded:
>>> 
>>> - that we didn't have a /8 to spare
>> 
>> - Allocating a /8 to ISPs, while it is still feasible, would only slightly modify the date
>> at which IPv4 prefixes er exhausted, not a big deal.
>> 
>>> - that a /8 wouldn't be enough, anyway
>> 
>> - ISPs already operate several parallel instances of 10/8 clouds, so that, although
>> a /7 would be more generous than better a /8, a new /8 is enough to replace
>> current 10/8's by a prefix private sites don't uses internally.
>> 
>> The known reason why ISPs have problems if they assign 10/8's is that some NAT44s:
>>  . assign 10/8's for internal use
>>  . cannot work with identical internal and external address spaces
>> 
>> CONCLUSION:
>> Assigning a new /8 to ISPs so that they can replace 10/8:
>> - is easy to do
>> - has negligible effect on the transition schedule
>> - does solve a real problem
>> It should, in view of this analysis above, be highly recommended by IETF
> <snip>
> 
> I'm still pretty young compared to most here, only close 15years in the Internet
> game but I've already seen enough to be scared of suggestions like the one
> that started this.
> 
> You're conclusion and the suggestion only make the matter worse really.
> You put of the problem into the future instead of solving it.
> If the problem is as bad as everyone say, why not make it 50 x 10.0.0.0/8
> instead of 40 x ? (someone said something about verizon had used over 40).
> You have the adresses, you know how to do it, just do it _and_ at the same
> time deploy IPv6.

The problem statement is NOT we are out of 1918 address space.  Please re-fresh your memory of the draft.  The problem is that the customer base uses pretty much all of 1918 address space on their side of their NAT gateways, I can not find a block that I can guarantee would not conflict with a customers use of 1918.  If Murphy strikes (and he will) and I assign, say 172.16.32.5/24 to the upstream interface of a CPE, with the default router of 172.16.32.1/24; and the customer has configured his side of the CPE to be 172.16.32.254/24, the routing stack in the CPE will become hopelessly confused.  Which interface to send traffic to 172.16.32.1?  Both are valid connected routes.

> 
> 
> When you've painted yourself into a corner, why also point the walls around
> you just because you are in a corner?
> 
> 
> -- 
> 
> Roger Jorgensen           |
> [email protected]          | - IPv6 is The Key!
> http://www.jorgensen.no   | [email protected]
> _______________________________________________
> OPSAWG mailing list
> [email protected]
> https://www.ietf.org/mailman/listinfo/opsawg
> 

---
李柯睿
Check my PGP key here:
https://www.asgaard.org/~cdl/cdl.asc

_______________________________________________
OPS-AREA mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/ops-area