On Sun, 2005-07-03 at 13:46 +0000, John Levine wrote:
If there is a possibility of two RR appearing in a resolver due to a
record modification and some overlap in the TTL of old to new, then:
There isn't. Somewhere in the std and bcp documents it says that all
of the RRs in an rrset must have the same ttl exactly to prevent this
kind of situation.
I was thinking of a change in targets as being a source of the problem.
With there being a restriction on only a single RR allowed in the set,
relative RR TTLs would not be an issue. The question is really whether
a change to this RR may cause a second RR to appear for some period of
time in a local cache. The TTL in the cache would be independent of the
new RR just published where there could be some overlap. By allowing
multiple RRs this situation, this would never become a problem. With
the restriction, a problem may occur when two RRs become cached for some
reason during a change. I have not experimented in creating this
problem. In this case, a temp error would prevent it from alerting
administrators without need. If the error does not resolve because
there ARE two being published, then this would become apparent
eventually, and should be a rarer case.
-Doug