does type forward zones make BIND 8 continuously poll?

[Cricket Liu]

> I'd like to disagree slightly with Cricket here,  BIND 8 prior to 8.23
> would query the forwarders in listed order, timing out at 30 secs
> divided by the number of forwarders.  BIND 9 times out in a different
> way which I find almost indescribable!  I've got some traces lying
> around somewhere if anyone's really interested, I'll forward them
> individually since IIRC Listar strips off binary attachments.

Hmm.  I've been reading the code and it looks like you're right
about the timeouts.  In BIND 9, the timeout when querying a
forwarder seems to be no different than the timeout when querying
any other name server.  In fctx_setretryinterval(), the retry timeout
is two seconds for the first two tries, as long as the name server's
RTT isn't larger than two seconds.  If it is, the retry is the RTT.
Oh, and the retry is always trimmed to 30 seconds.

However, Bob's comment in fctx_getaddresses() certainly suggests
that the forwarders are queried in a fixed order:

                 * XXXRTH  We could sort the forwaddrs here if the caller
                 *         wants to use the forwaddrs in "best order" as
                 *         opposed to "fixed order".
                 */

And, sure enough, sort_finds() sorts fctx->finds rather than
fctx->forwaddrs, which I believe means it's only sorting the list
of name servers we've found in our local database while doing
iterative name resolution, not the forwarders.

I'm copying bind9-users on this, since most of the BIND 9
developers hang out there.  They can give us the authoritative
answer.

cricket

Men & Mice
DNS Software & Services
www.menandmice.com

Attend our next DNS and BIND class!  See
http://www.menandmice.com/8000/8000_dns_training.html
for the schedule and to register for upcoming classes