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

-----Original Message-----
From: [email protected] [mailto:[email protected]] On Behalf Of 
Victor Kuarsingh
Sent: Thursday, November 11, 2010 2:29 PM
To: james woodyatt; IPv6 Operations; [email protected]; [email protected]
Subject: Re: [v6ops] [OPSAWG] Heads up - Plea for allocating a /8 to ISPs

>We have found problems in deploying NAT444 with RFC1918.

[Wes]
I'm not going to pretend that this works perfectly, but be specific about the 
problems you've found. Not a general discussion of what NATs or multiple 
layers of NATs break - specifically what does using RFC1918 as a CPE NAT 
device's external interface break? That needs to be in the draft to contribute 
to the technical justification for this proposed allocation. Or, it needs to 
be in http://tools.ietf.org/html/draft-donley-nat444-impacts-01 with a 
reference to it from this draft.

The one that I'm aware of is the overlap between internal and external address 
blocks (i.e. they can't be the same). Regarding this, the quote I heard (from 
someone, can't remember whom) was, "we have tested this and we can find at 
least one CPE device that uses every address block within RFC1918 on their 
internal network segment".

I would like to see some hard evidence to back up this claim, and at the /24 
level, not at the supernet level. I assume it's available if it's been tested.
I think that what you'll find is that if you line the spread up based on 
number of devices and deployment numbers within each network, there will be a 
bell curve, and the solution is actually to avoid the most common ones and 
manage the breakage on the ones that are out at the long tail. I think it's 
acceptable to have a fractional percentage of customers that need to have a 
tech support person talk them through reconfiguring their router to use a 
different network block to fix things, or even replacing their customer-owned 
router with an ISP-owned (and hopefully IPv6-capable) one. As Joel and the 
draft above said, the other reality here is that NAT444 is *going* to break 
things. Why is it verboten for it to break a small percentage of CPE because 
the vendor in question decided to be different and use 10/8 as its default 
internal network?

Wes George

_______________________________________________
v6ops mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/v6ops