Re: dnscache querying wrong (old?) NS sometimes

> On 9/11/2009 12:57 AM, Matthew Dempsky wrote:

> I think the problem Rihad is referring to is when:
> 
> 1. example.com has NS to [ab].ns.example.net
> 2. his dnscache knows things about example.com
> 3. the example.com owner changes NS to different servers
>     but did not follow Step 4 in
>     http://cr.yp.to/djbdns/run-server.html#move-zone
> 4. dnscache continues to ask [ab].ns.example.net for data even
>     though the gtld servers are handing out different authoritative
>     servers.
> 
> The solution is in Step 3 -  dnscache will then query [ab].ns.example.net 
> for example.com and either find they are no longer authoritative and/or 
> will find new NS data - forcing a fresh lookup.
> 
> see http://securepoint.com/lists/html/djbdns/2007-01/msg00033.html
> 
> I agree with Charlie Brady (in the thread) that this is a bug in the dns 
> rfcs, and that BIND's position on the topic is preferred over dnscache's. 
> But i dont think anyone's come up with a patch to change the behavior.
> 

Wow, perfect, you've just read my mind. How ugly... As I'm not competent 
enough to hack dnscache, I think I'm forced to go back to BIND 9's cache 
or some other. This issue comes up once every week or so, and will 
surely only worsen as our ISP grows.

DJB is a great contributor, and I think we should all thank him for 
sharing the code with us, but yet he is very stubborn with respect to 
"RFC bug is not my bug" ;-) Software not only has to be reliable, but 
also has to be usable.