LIMITATIONS IN FOLLOW

Follow 2.x does make many assumptions about both the HTTP, a Web server, and
the surfing habits of your users. While they should give a good indication
of how your site is used, there are situations, both on the server and
client side, that will be misinterpreted by Follow. As time goes by, Follow
will become more refined, but there are some things that simply cannot be
found out from a logfile.

First and foremost, Follow is dependant both on what the Web server writes to
the logs, and what the browser tells the server. Follow was developed with
Apache 1.2.x, and may not operate as accurately with other servers.
Likewise, what is in the referer field is completely dependent on the
clients' browsers; although extensive testing of a variety of browsers and
platforms has taken place, there are doubtless special conditions that are
not accounted for.

In particular, if a user is on a Web page when they type in a manual URL (or
use a bookmark), many browsers will send that page's URL as the referer.
This causes a degree of background noise that must be taken into
consideration.

Likewise, a logfile gives a very small window into what is happening on a
user's desktop; they may have multiple browsers open. Because of the very
open nature of the Web, there are almost limitless possibilities of client
actions and behaviors; some will confuse a program like Follow.

On the server side, CGI scripts can take many forms, and the logs may not
faithfully reproduce what they're doing. As a result, Follow does not treat
them diffferently; it cannot tell what different CGI behaviors mean to the
client. Particularly troublesome CGI's can be ignored in follow.conf.

The calculation of BACK paths is reasonably accurate, but can be fooled, and
does not take into account the entire path, but only the page whose referer
matches a previous page.

Of course, frames Web sites present a challenge; currently, they are not
supported. 

One thing to keep in mind when considering these issues is that Follow is
interested in user tracking, not hit tracking. Many conditions that would
lead to confusion (such as multiple user machines, proxy caching, etc) will
produce 'BROKEN' hits. Sessions which include these can be filtered out in 
follow.conf, leaving a relatively 'clean' sample.

I hope that these explanations will help you understand what context Follow
should be used in.