At 11:59 AM 3/8/04, Greg Connor wrote:
I think it's correct that one record can have only one TTL, but in
Marc's example there are two MX records. I don't actually know what
happens at the cacheing server if one record expires and the other does
not.
I could see having two different TTLs, maybe if the secondary doesn't
change but the primary might change often. But, I would probably want
to test this out, or else use the lower ttl for both to be on the safe
side.
Also, you can have a long TTL time for the MX and a short TTL for the
corresponding A records, which would allow you to change the IP address
quickly if you need to.
--"Arnold K." <aksup(_at_)voicenet(_dot_)com> wrote:
Greg,
The example deals with an MX record.
i.e.
domain TTL IN MX is the record.
The preference and server are instances within the record.
You're right that you can have one TTL for the MX record and a different
TTL for the A record.
OK, I swear that I have seen RFCs refer to them as multiple records, or
multiple RR's (resource records). An answer (or answer section) can
contain multiple records. I don't believe I have seen the word "instance"
used before.
However, despite the semantic differences, after a couple of brief tests it
seems you are right about the TTL... multiple records of the same type,
with the same label, will be assigned the same TTL (in the case of BIND it
was the first one given).
--
Greg Connor <gconnor(_at_)nekodojo(_dot_)org>