DNS cache issue

Danny Mayer mayer at gis.net
Sun Nov 25 00:16:19 UTC 2007


Adam Tkac wrote:
> On Thu, Nov 22, 2007 at 12:46:18PM -0500, Paul Wouters wrote:
>> On Thu, 22 Nov 2007, Adam Tkac wrote:
>>
>>>> atkac at redhat.com:
>>>>> Of course, EDNS is enabled by default
>>>> I see a contradiction here. If it's enabled by default in the Red Hat
>>>> version, there shouldn't have to be a "MUST add" to enable it.
>>>>
>>>> So, which way is it? :-)
>>> If you don't specify edns option at all you have EDNS enabled. If you
>>> specify "edns yes;" you have same behavior but your named.conf is
>>> longer :)
>> I ran into issues where named told me "too many EDNS failures, disabling".
>> Then I started looking around and found the edns option. I set it to
>> yes, and the errors disappeared.
> 
> It's mystery that you have different behavior with 'edns yes;' and
> without edns option.
> 
>> I guess it might have been something else, if edns is enabled by default?
>>
>> Also, I am still curious why you turned this into an option. Is there
>> any experience you had where you needed to disable edns?
> 
> Yes. I had some reports in our bugzilla that syslog is flooded with
> messages about disabling EDNS. I believe that problem is in broken
> routers. But for many people is far more easy disable EDNS than report
> problem to their router vendor. Global ends option is for that people.
> 
> Adam

Doing the easy thing doesn't make it the right thing and just does a
disservice to your customers as well the router vendors not being told
that they have to fix these problems. Ignoring something doesn't make it
go away and soon they will have problems that won't be ignorable in this
way.

Danny
> 
>> Paul
> 



More information about the bind-workers mailing list