Force TTL on a Caching Name Server

[Chris Thompson]

On Sep 7 2006, Mark Andrews wrote:

>[Kevin Darcy <kcd at daimlerchrysler.com> wrote:]
>
>> howting at gmail.com wrote:
>> > I got it working by adding the following 2 lines to the options block:
>> > max-cache-ttl 300;
>> > max-ncache-ttl 300;
>> >
>> > After restarting named, dig shows that all TTL are capped at 300 :)
>> >
>> > This is the perfect solution I wanted. Thanks for all the help!
>> >   
>> Hmm... OK. So if one of your users really likes Hemis (TM) and hits 
>> http://whatcanyouhemi.com frequently during the course of a day, you 
>> think it's "perfect" for your nameserver to be querying mine 12 times as 
>> often as it should be expected to (the TTL I've set on the RRset is 1 
>> hour, but you're capping it at 5 minutes)? That's an interesting 
>> definition of "perfect". It's certainly not "perfect" from the 
>> perspective of my traffic load, my capacity planning for the future, my 
>> budget, etc.
>> 
>> The DNS infrastructure of the Internet is held together in large part 
>> through mutual consent and co-operation. Practices such as you are 
>> describing undermine those, inasmuch as you are putting a burden on 
>> others for your own selfish reasons. I would seriously urge you to 
>> reconsider.
[...]                                                     
>
>	And any competent C programer should be able to add the two
>	or so lines it would take to cap the ttl in responses w/o
>	having to effectively kill the cache.

It isn't entirely clear that the OP wants only to limit the TTLs in
downstream responses. But if that is the case, then he could use two
caching nameservers: A which is uncrippled and talks to the outside
world, and B which has the max-*-ttl settings and forwards all its
requests to A.

Saves writing those "two or so lines [of C]" ...  :-)

-- 
Chris Thompson
Email: cet1 at cam.ac.uk