Re: [secdir] secdir review of draft-ietf-dnsop-as112-dname-04

[Paul Wouters <[e-mail]>]

On Thu, 7 Aug 2014, Klaas Wierenga (kwiereng) wrote:

> For security considerations relating to AS112 service in general, see RFC6304bis]."
>
> And in the introduction
>
> "The AS112 project does not accommodate the addition and removal of
>   DNS zones elegantly.  Since additional zones of definitively local
>   significance are known to exist”
>
> It appears to me that moving from a static list of well known ranges as defined in 6304bis to a dynamic adding and dropping of IP-address ranges that are "of definitively local
>   significance” is prone to error, potentially leading to blackholing valid address space, either on purpose or by accident. I may show my ignorance here, but I feel that a discussions as to why that risk doesn’t exist is warranted.

Actually, the static list is much more problematic. Mostly because some
AS112 instances would not yet have the new addition, while other AS112
instances would, and the replies would be inconsistent. By using the
DNAME protocol, it makes anything thrown at an AS112 work. Anyone else
throwing something in would still cause less packets on the internet
than those DNS packets leaking out the regular DNS infrastructure. For
instance, ISPs could run an AS112 instance to take all RFC1918 queries.
When new ranges are added (like 100.64.0.0/16) the ISP does not even
need to update their AS112 instance (they have to do so now). The
document's main purpose is to get rid of that bad static list. Because
it is static, an ISP cannot even really add anything to it, without
causing inconsistency.

The only people that can make a mistake to "null route" their domain into
AS112 are the owners of the domain. Don't configure NS records that
point to AS112 and you are fine.

Paul