Diffrent DNS Answer by diffrent IP Source Adress

Kevin Darcy kcd at daimlerchrysler.com
Wed Aug 16 01:39:31 UTC 2000


Jim Reid wrote:

> >>>>> "Kevin" == Kevin Darcy <kcd at daimlerchrysler.com> writes:
>
>     Kevin> Another approach, which may meet your needs, is to add both
>     Kevin> A records and sort the responses via the "sortlist"
>     Kevin> mechanism of BIND. This gets you out of the business of
>     Kevin> maintaining multiple copies of zones, but it does require
>     Kevin> that *every* nameserver answering the name be configured
>     Kevin> with these sortlists, which can be a logistical
>     Kevin> nightmare...
>
> It's also not scalable or flexible. How many A records and network
> masks would someone want/need to sort?

IIRC, the original poster indicated 2 different networks. I'm using
sortlists for a much larger number than that (I haven't delved much into
our A class yet, but I've already sliced heavily into one of our B
classes).

> What if it was the A record for
> a web server that was to be sorted, but not any other A records on the
> same net as that web server?

Theoretically, the second element of a sortlist statement can be as
narrow as you want, even just single IP addresses (you can call them
/32's if you want :-). Admittedly, my experience with sortlists so far
has been of the simplistic BIND4-style "network sorting" variety, but the
mechanism has more flexibility than that.

> IIRC the original poster wanted *different* addresses to be returned
> based on the source address of the query. Using a sortlist{} clause
> would return the same set of addresses, albeit in a different order.
> This might or might not meet the original poster's requirements.

Agreed. Which is why I prefaced my response with "Another approach, which
may meet your needs, ...". I was fully cognizant that sortlists didn't
meet his stated *requirements*, but given what a pain in the ass split
DNS is to set up and maintain, I figured an alternative was worth
mentioning.


- Kevin




More information about the bind-users mailing list