From: Adam Shostack <adam@lighthouse.homeport.org>
Date: Thu, 15 Feb 1996 19:28:06 +0800
To: accessnt@ozemail.com.au (Mark Neely)
Subject: Re: Off topic - research query
In-Reply-To: <199602140649.RAA14008@oznet02.ozemail.com.au>
Message-ID: <199602141518.KAA25093@homeport.org>
MIME-Version: 1.0
Content-Type: text


Mark Neely wrote:

| As a side issue, I wanted to cover the "overhead"
| factor inherent in the TCP/IP (v4?) protocol 
| which I understand is reduced under the proposed IPv.6 protocol.

	Comer is the best text on IP, his third edition talks about
IPv6 as well as v4.  Essentally, there was a some unneeded stuff in
IPv4 headers, which routers had to look at.  The IPv6 headers are much
cleaner.

| I'd also like to discuss the "unfriendly" manner in which
| web browsers such as Netscape hog resources by sending multiple
| port access requests.

	Our own Simon Spero has a paper entitled something like
'Optimizations for HTTP.'  Its on the w3.org web server.

	Speaking of HTTP-ng, I was thinking the other day about a
scheme for further optimization.  It only works in the presense of
dnssec, which is moving forward.

	When getting a URL, add a meta tag, which gives the web
server's idea of what the referenced hosts IP address (or its primary
NS) is.  This could be a win because we need fewer calls to the root
name servers.  Those calls tend to be short, (1 or 2 udp packets each
way), which need to be routed in a way that few other packets would
need to be.  By eliminating them, we push all of the traffic regarding
a web host to its network.

	This only works under DNSsec because otherwise I could say
http://www.microsoft.com/ meta-dns-A=140.174.1.3

Adam

-- 
"It is seldom that liberty of any kind is lost all at once."
					               -Hume