Is your browser taking forever to connect to the world? Try this.

[reprobate]

It was taking literally over a minute for Firefox or Safari to connect to anything that wasn't cached. A little research - really only a couple of hours - and a suggestion from the Apple discussion forums put me on the right track.

Seems like Apple, in the switch to Leopard, slightly changed the way a request is sent to the DNS server, resulting in the effect that those servers that were up to date with the specifications were not a problem. Those that were using old code would time out and the secondary server would take over resulting in loooong wait times.

Since the hold up was only in the "looking up" period, I figured it had to be a DNS holdup and took the advice of the poster on the Apple board, switching from my Roadrunner DNS to the servers at dns.com. This is easy to do on the Apple Airport Extreme that I'm using. Just bring up Airport Utility, select the router you use to connect, manual configure, then internet. The DNS server numbers are about mid page. Just click on them and enter the new server numbers, then click update. That's it. Of course if you're using a different router your method may vary.

Connections to web pages are now almost instantaneous, with the longest wait taking up to three or four seconds. Frustration quotient dropped to acceptable levels.

[WritersBlock]

1. dns.com or opendns.com?

Went to dns.com and it looks like a placeholder... did you mean opendns.com, maybe? :)



On edit... we tried opendns.com's nameservers, and it has made a big difference. Thanks for the tip!

[Pierre.Suave]

2. Interesting

[moggie]

3. Very vague

What, specifically, is different about DNS queries from Leopard, and what's the difference between the way the Roadrunner and dns.com servers respond? Is this related to queries for IPv6 addresses? I've done a quick trace of traffic resulting from Safari on Leopard, and I see it sending AAAA (IPv6 address) queries as well as the more usual A (IPv4 address) queries. Or is it something to do with SRV queries? I've seen that mentioned, but a quick check on 10.5.3 doesn't show that behaviour.