At 11:54 AM 8/5/98 -0400, you wrote:
>On Wed, Aug 05, 1998 at 10:34:00AM +0100, Philip Hazel wrote:
>> How often is occasionally? I don't myself think this is a major issue,
>> but others may wish to argue otherwise.
>It may be overly complex but why not use the Time-To-Live value associated
>with the address in the DNS? After the TTL expires, an address is not
>guaranteed to be valid anyway, so that would be a good time to reload.
yes!
while TTLs are often up in the 24 hour range, when i transition a domain
name to a new ip address, i start shortening the TTLs a day in advance -- i
cut it to one hour, then, 24 hours later, i cut it to a small number (say 5
minutes), and then an hour after that, i can think about a reasonably clean
transition to the new IP.
in other words, TTLs do get manipulated intentionally for good reason.
but here's another issue: i always run a caching name server on any
mailhost that sees serious service. since it already implements all the
caching mechanisms per RFC, perhaps there should be an option:
ip_cache = no
ip_cache = yes
richard
richard
--
Richard Welty
NeWorks Networking, Inc. 518-244-9675
rwelty@??? http://www.neworks.net/
--
*** Exim information can be found at
http://www.exim.org/ ***
