Re: Robots

From: Jim Pitkow (pitkow@parc.xerox.com)
Date: Fri, Dec 25 1998


Message-Id: <4.1.19981224202212.0097e8a0@mailback.parc.xerox.com>
Date: Thu, 24 Dec 1998 21:02:07 PST
To: Henrik Frystyk Nielsen <frystyk@w3.org>, www-wca <www-wca@w3.org>
From: Jim Pitkow <pitkow@parc.xerox.com>
Subject: Re: Robots

At 05:07 PM 12/13/98 , Henrik Frystyk Nielsen wrote: 
>
> Yes, this was what I was trying to get at :) Central registries are good for
> some things and less good for others. One way to decentralize things would be
> to let the user-agent field be a URI which could be dererenced in order to
> get information about the application. 


It be nice not to have to access the Web to decode the URI field, as with the
problems of reverse DNS lookup for active servers that only log IP addresses. 
I'd like to see the following fields:

Browser         Netscape, Internet Explorer, Opera, WebTV, etc
Browser Version 5.01-a, etc.
Platform                Win 95, Solaris, etc.
Platform Version        sr2, 2,7, etc.
Autonomy                User, Semi-Autonomous, Robot (maybe as UIR?)

>
> However, as the same application may be used for different things, this is
> not always good enough. This is why I hope it is possible to deduce the
> "access mode" by looking at the logs. 


I did some work back at Georgia Tech that characterized user browsing behaviors
based upon the length of their surfing paths.  Linda Tausher did this at the
University of Calgary for her MS.  We've done this with the Law of Surfing work
as well.  The problem with these methods though is that the cognitive states
induced from log files may not actually correspond to what users are doing. 
What is really needed is think aloud protocols, video taped session, and
in-depth questionnaires to figure out what users are doing.  Stated
differently, no one has yet to show that the behavioral patterns deduced from
log files really corresponds to what users are doing. 

Jim.