At 8:17 AM -0800 11/14/08, Tony Finch wrote: > >Note that I'm not arguing against a new RR type, I'm just trying to >understand the arguments against the de facto standard. > >One significant advantage which I have not seen clearly articulated is >that a new RR type could combine the functions that are currently >performed by A records (bit vector) and TXT records (explanatory URL) >which could greatly reduce the number DNS lookups. > I don't think is really the right list to go into this in detail, but I agree that there are potential benefits to shifting to a new RR beyond those related to the coherence of the DNS model. Depending on the syntax, it could well give you ways to distinguish among cases which I believe some DNSBLs conflate now. For example, giving you ways of distinguishing among the following cases: 1) We have no data, positive or negative, about the record queried. 2) We have data about the record queried and it has a positive reputation 3) We have data about the record queried and it is mixed; we see spam and we see non-spam email as well. 4) We have data about the record queried and we believe to be all spam 1 & 2 are easily conflated, as are 3 and 4. Giving you richer ways of handling that would enable you to let the customer make more individualized decisions. Data on the freshness of the information is also not easily carried in the A-record style response (without then re-using another aspect of the DNS, the TTL, in a subtly different way). Many users may not want to interpret this data, obviously, as they want the simplest check possible so that run-time processing is possible. But it is trivially easy to re-conflate. regards, Ted Hardie _______________________________________________ Ietf@xxxxxxxx https://www.ietf.org/mailman/listinfo/ietf