Re: Persistent Documents and Locations

Keith Moore (moore@cs.utk.edu)
Mon, 21 Aug 1995 20:13:35 -0400


Message-Id: <199508220013.UAA04304@wilma.cs.utk.edu>
From: Keith Moore <moore@cs.utk.edu>
To: lazear@dockside.mitre.org
Cc: "Karen R. Sollins" <sollins@lcs.mit.edu>, uri@bunyip.com, moore@cs.utk.edu
Subject: Re: Persistent Documents and Locations 
In-Reply-To: Your message of "Mon, 21 Aug 1995 15:41:41 EDT."
             <9508211941.AA21609@dockside.mitre.org> 
Date: Mon, 21 Aug 1995 20:13:35 -0400

> This why cache schemes need timeouts, so that people don't
> believe forever the information they've gathered.  So, how
> long should a URL be valid for?

It varies considerably from one URL to another.

One problem with DNS TTLs is that they are usually picked "out of thin
air", with very little consideration as to how likely it is to change.

Also, no matter what the DNS says the TTL is, the change to the DNS is
made with no warning.  Often there is no provision to handle network
services from remote clients who are using old DNS records from their
caches.

My point is: TTLs aren't sufficient. You need a system that can move
things from one place to another while continuing to maintain the old
copy of the file until everyone's cached records have expired.

Keith