rDNS Round-Robin

Kevin Darcy kcd at chrysler.com
Wed Jul 22 17:50:38 UTC 2009


Matus UHLAR - fantomas wrote:
>> On Wed, Jul 8, 2009 at 5:08 PM, Mark Andrews<marka at isc.org> wrote:
>>     
>>>        RRsets are unordered.  Software and configurations should
>>>        be prepared for this.  Where ordering is required it is
>>>        built into the RR type.
>>>
>>>        Mark
>>>       
>
> On 14.07.09 14:02, Bryan Irvine wrote:
>   
>> I've think I've found the confirmation I was looking for in RFC 2181
>> section 10.2.
>>
>> Does this seem to confirm that round-robin PTR's are perfectly legal?
>>     
>
> yes, they are perfectly legal. However I don't know about any application
> that would require nor benefit of them, and I don't recommend using them.
> With most of applications doing reverse resolution and using its result
> anyhow it's still better to have always the same name...
>   
Since we're nitpicking standards here, let's be clear that there is a 
distinction between "multiple-record RRsets", which refers to the 
structure of the DNS database in a particular area, and "round-robin", 
which refers to how a multiple-record RRset is treated when being given 
in a response from a resolver.

It is perfectly legal to have multiple PTR records in a given RRset.

It is also perfectly legal for a resolver to "round robin" the records 
of a PTR RRset in its responses.

"Round robin" behavior is not, by standards, *required* of any resolver. 
As Mark put it "RRsets are unordered".

So, if the question is: "does a round-robin PTR conform to standards?", 
then the answer is "yes". Both elements of that -- the RRset having 
multiple records and the resolver performing 'round robin' sorting of 
those records -- are optional and legal.

But, if the question is: "given a PTR RRset with multiple records, 
*must* this result in a 'round robin'?" then the answer is "no". No 
resolver is *required* to "round robin" anything. If it gives the RRset 
always in a "fixed" order, or randomly, or using some other algorithm, 
e.g. optimizing the response to place addresses that are considered 
"closer" to the requesting client at the top of the list (BIND does this 
via its "sortlist" facility), then these are all legal.

Hopefully that clarifies things.

                                                                         
                           - Kevin




More information about the bind-users mailing list