Re: [DNSOP] Mitigation of name collisions

["Paul Hoffman" <[e-mail]>]

On 18 Sep 2016, at 15:21, John R Levine wrote:

>> It is impossible to measure the effectiveness without knowing how 
>> many collision queries are just noise (queries that will cause no 
>> noticeable damage if they started coming back with results).
>
> Agreed.  I don't see how to find that out in ways that are not hard to 
> back out if it turns out the damage is as bad as we fear.

I do see that, but that's a discussion for a different time and place. 
(Unless this WG re-adopts corp/home/mail, of course.)

>
>> In the case of mitigation through wildcard-to-localhost, it is safe 
>> to assume that many organizations did in fact mitigate; we simply 
>> can't tell how many or when.
>
> How come?

Because a few of them told me they did.

> I'm not denying it's possible, but I've never seen any evidence that 
> there were collisions to mitigate.

You of all people should know that "people do dumb things with the DNS". 
:-)

> Before the 127.0.53.53 approach, some TLDs tried reserving the names 
> that showed up in DITL snapshots, and those names looked to me totally 
> random, likely generated by something that was trying to see whether 
> some piece of namespace was wildcarded.

As we saw at the collisions workshop 
(https://www.ietf.org/id/draft-thomas-namecollisions-workshop-report-05.txt), 
DITL data is poorly suited for following collisions because you can't 
tell how much is coming from organizational resolvers that are in front 
of a poorly-chosen name and how many are from upstream ISPs.

>> (Disclaimer: I'm now on ICANN staff, but well before I was, I wrote 
>> "Guide to Name Collision Identification and Mitigation for IT 
>> Professionals" for ICANN.)
>
> A fine document for people who already realize they need to deal with 
> collisions, not so much for people who don't realize they exist or 
> assume they're someone else's problem.

Correct. It has been helpful, though, at least to organizations seeing 
127.0.53.53.

--Paul Hoffman