Re: "Top N misconceptions about Web traffic that Internet implementers should know"

From: Balachander Krishnamurthy (bala@research.att.com)
Date: Mon, May 24 1999


Message-Id: <199905241655.MAA51732@raptor.research.att.com>
To: "Martin F. Arlitt" <arlitt@hpl.hp.com>
Cc: www-wca@w3.org
Date: Mon, 24 May 1999 12:55:47 -0400
From: Balachander Krishnamurthy <bala@research.att.com>
Subject: Re: "Top N misconceptions about Web traffic that Internet implementers should know" 


i would be curious to know more about how you tested the servers:
if nat was being used, if OPTIONS returned the same response
each time, if you tried to pipeline, if you tried any of the other
myriad 1.1 features, is the "1/2" that supported persistent connections
is a reference to unique "1/2" of servers (i.e. IIS, Netscape Enterprise,
Apache 1.3.4 etc.) or just half of the ones you contacted etc etc.

we have to be very careful not to fall in the same tar pit as netcraft
and other "surveys" which talk about penetration of 1.1.

speaking just for myself, i would be very very skeptical of any and 
all such surveys that are done without a sound understanding of 
end-to-end vs hop-to-hop info, proxies along the path, uniqueness
and idempotency of servers being contacted, reliance on version numbers
of servers etc.

martin: you and i can take this offline and do a in depth experiment if you
        would like. 

bala