2 ms·
> most reputable sites are not changing all their IP's very often Your suggestion is that only disreputable sites are changing their IPs frequently, but there
by jodrellblank 17y ago
> most reputable sites are not changing all their IP's very often
Your suggestion is that only disreputable sites are changing their IPs frequently, but there are a lot which return different addresses for load balancing or geolocation or CDN reasons, or which genuinely might change (MX records particularly) quickly, and some which are generated on the fly so never-typed-before.example.org will resolve, because they've configured a server to reply to any request but there is no inclusive list to cache.
> if (and only if) they do, then we can requery
And how will the client know? If the site is down it might be down for real or the client might have no internet connection, instead of the IP being outdated. If there's another site on the old IP then it will connect and return valid headers. You'd have to rely on the user already knowing the website they wanted to see and saying "that isn't right try again but properly this time" and that assumes it is something user facing rather than a POP3 connection which suddenly "can't connect" or gives a login failure.
Plus you'd then need every popular TCP/IP using program ever to be rewritten to have a "try again but this time with a full DNS requery" option.
> 50+ queries to a varied array of servers to find the IP's for one host is not unusual
So fix that, don't overhaul the whole system because some people can't configure it properly. As you said it shouldn't need more than three or four at most. Where are you seeing that so often that you say it is "not unusual"?
> so ultimately, if we follow the path to the 'top', the whole scheme relies on people trusting people
As does the whole SSL certificate pyramid scheme and the whole 'using an OS that was written by other people' scheme and the whole 'eating food grown and processed far away' scheme. That doesn't mean it's inherently awful.
> we have to trust the telecom companies for the directories they publish.
And we have to trust them to route packets to their destinations once we have the resolved IP as well and not change the content on the way over, too.
> popular browsers and OS's are themselves caching IP's mostly outside of the user's control, then are these load balancing devices irrelevant?
No, they load balance for the site over many users, not between individual requests from the same user. And if OSes and browsers are caching why bring the hosts file into it?
> the HOSTS file. even packed with 1000's of IP's, it's relatively small. What are your compelling reasons?
You can run your own caching name server (on your computer or LAN) right now, you don't need to overhaul DNS for the whole world in order to benefit. Plus, local hosts file management generally would be a pain in the neck.