Re: How long does an IP address change take?

[Jordan Bettis <jordanb@hafd.org>]

On Thu, Nov 01, 2001 at 02:53:43PM -0600, Jay Link wrote: 
> How do I know this? Well, a) I've read about it, and b) having done some
> web pages for people on AOL, I've experienced first-hand how they often
> cannot see changes immediately, and sometimes for days. Although, I would
> have guessed 72 hours was the max on that, too.

I've used Squid as a caching web proxy before. I'd assume AOL has something
similar. It can normally figure out what can be cached and what can not, 
which static pages have changed since they were last looked at, etc. It
presumably does this the same way your web browser does it. 

I've seen Squid get confused before where the only way to see an updated
page is to wipe squid's copy and force it to download a new one, but that
is rare. AOL's caching servers probably have that problem from time to time
as well. 
 
> Robert, is it possible that we're also dealing with the cache on the
> client's computer? That is, could their browsers be set to "never update",
> and could the problem be solved by them simply refreshing their display?

About the DNS problem. DNS is designed at the protocol to be a caching system, 
it reduces bandwith requirements and loads on popular servers (like the 
ICANN roots). You set the TTL value in the zonefiles. Next time you make
a change, reduce the timeout value beforehand.

You can set the TTL (in seconds) by puting it in the record, for example:

foo.example.com 86400 IN A 192.168.0.0  

sets the TTL to 86400 seconds, or one day. Since TTL is an optional field,
if it is ommited, the TTL defaults to the last explicitly stated TTL. The
standard isn't clear what the TTL should be if there are no TTLs defined
in the file, so your nameserver probably has an internal default.

-- 
Jordan Bettis <http://www.hafd.org/~jordanb>
The only constructive theory connecting neuroscience and psychology will 
arise from the study of software.
          -- Alan J Perlis: Epigrams in Programming, ACM SIGPLAN 1982
-
To unsubscribe, send email to majordomo@luci.org with
"unsubscribe luci-discuss" in the body.