Re: [DNSOP] Mirja Kühlewind's No Objection on draft-ietf-dnsop-dnssec-roadblock-avoidance-04: (with COMMENT)

["Mirja Kuehlewind (IETF)" <[e-mail]>]

Hi Wes,

thanks! See below.

> Am 09.07.2016 um 00:28 schrieb Wes Hardaker <[e-mail]>:
> 
> "Mirja Kuehlewind" <[e-mail]> writes:
> 
>> 1) Shouldn't/can't section 3.1.13. (UDP size limits) also specify a
>>   real test?
> 
> I don't think it's possible to easily test this, sadly, without a target
> set containing different deterministic response sizes.  We could
> probably strike the section and simply state that resolvers MUST support
> TCP (which actually is stated elsewhere). 

In this case it probably make sense to not have an own section but mention somewhere else that the packet size can have an influence.

> 
>> 2) To follow up with the tsv-art review: To avoid network as well as
>> server overload, would it be useful to provide further guidance, when and
>> how often these tests should be performed?
> 
>      <t>These tests should be performed when a resolver determines
>      its network infrastructure has changed.  Certainly a resolver
>      should perform these tests when first starting, but MAY also
>      perform these tests when they've detected network changes
>      (e.g. address changes, or network reattachment, etc).</t>
> 
> FYI, I don't think even with a lot of boxes starting at the same time it
> would cause significant overload.  Specifically, those resolver boxes
> are serving many more clients that will be issuing significant more
> traffic once the resolver is operational than these tests actually
> require.

Thanks text sounds good. I’m not worried but it always better to give some guidance otherwise the weirdest implementations might show up.

> 
>> 3) In section 6.1.  (What To Do): maybe also list logging as an option in
>> cases where no user is directly involved but a human might check later.
> 
> Good point.  I've changed it to:
> 
> 	<t>If Host Validator detects that DNSSEC resolution is not
> 	possible it SHOULD log the event and/or SHOULD warn user. In
> 	the case there is no user no reporting can be performed thus
> 	the device MAY have a policy of action, like continue or
> 	fail. </t>

Great! Thanks!

Mirja


> h
> -- 
> Wes Hardaker
> Parsons
>