On 24/Jan/12 22:34, Murray S. Kucherawy wrote:
>> From: John Levine
>> 
>> The next question has to be: if you have an external source telling
>> whose signatures to report, why wouldn't that source also tell you
>> where to send the reports and how many to send?
> 
> Wait, isn't that what I had in the last version, where the
> reporting address was in the key record in the DNS?

Not exactly, John suggests a separate record like:

 _report._domainkey.blackops.org TXT "r=sendreportshere"

Scott notes this solution allows to notify missing or misspelled
selectors.  So, as long as the ADMD don't change both records
simultaneously, this way provides for more stable reporting than
otherwise.  In addition:

 * "_report" is not a valid selector, so there's no conflict, and

 * for a range of [report data] X [key data] sizes, this solution
   may allow to use UDP rather than TCP.

For the consistency problem, and also in order to avoid redefining
report modalities at every turn, I make an appeal to treat them all
together, in reporting-discovery.
_______________________________________________
marf mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/marf

Reply via email to