Client

Server B

Server C

Server A
Client

Server B

Server C

Server A

Lamprey
Conventional HTTP transaction
Client talks directly to web servers HTTP transaction with Lamprey tracking Client talks through Lamprey CGI to web servers

Fig. 1: Lamprey system architecture compared with conventional web interactions.

Fig 2. Lamprey tabular log

Original URL:

<A HREF="http://site.host.edu/
projects/index.html">Our Project</A>

URL After Lamprefication:

<A HREF="http://www.lamprey.stanford.edu/ trackercgi?url=http://site.host.edu/pro jects/
index.html&user=smith@stanford.edu& dts=09:10:15">Our Project</A>

Once the user sends a URL to the Lamprey application, all subsequent activity is logged (until the browser window is closed). In our current use of Lamprey, we track user information by asking them to log in as part of our experiment, although this is not a technical requirement. The anchor text that the user sees is unchanged: for both URL anchors above, the user sees ?Our Project? as a hyperlink to click on. Only the underlying destination of the link is changed to send all URL traffic through Lamprey.

This process currently only tracks hyperlink and forms submitted across HTTP. Lamprey?s parser recognizes a subset of HTML 3.0[7] and assumes rigid adherence to this specification. URLs that use FTP, Gopher, S-HTTP and other protocols are left intact and not tracked. Lamprey does track forms and search requests, but does not track the data entered into password fields for restricted access pages.

Using this method of altering the HTML source, Lamprey tracks a number of statistics:

? The id of the user being tracked (typically an email address obtained through a log-in page).

? What page they are coming from, including when they arrived at that page.

? What link was activated (i.e. what page the user is going to), and when that link was activated.

? Additional client information such as what WWW browser they are using, and their IP address.

We have developed three methods of viewing Lamprey log files. All three methods are available through a WWW-based interface and can be viewed from remote locations.

Tabular report (Figure 2)
A formatted tabular listing of log entries is useful for examining the raw data at its finest granularity. A Lamprey tabular report includes a simplified version of the basic time-stamp data in the log file.

Footprints (Figure 3)
The ?footprints? report shows a formatted history of the user?s actions. The reporting mechanism performs abstractions on the log data to simplify the format. For example, although Lamprey does not directly track the usage of client-side interactions such as clicking on the ?Back? button (these actions are not