Testing and Evaluation

The testing of Rumour was done in two parts. The first part was to test out each component of Rumour, and the second part was to test the interaction between the Web servers that were using Rumour.

The first part of testing was done upon the completion of each component. These tests were used mainly to see if each component is doing what it is supposed to, and to see how they would react to strange inputs.

The second part of testing began when all the components worked satisfactorily after the first test. Rumour was installed onto the following Monash University Web servers :

• http://ip4.cs.monash.edu.au:8000/
• http://indy14.cs.monash.edu.au:8000/
• http://www.cs.monash.edu.au/
• http://yoyo.cc.monash.edu.au/

They were installed at my own directories, and the Webmaster at the above Web servers made a link to Rumour directories from the root document directory. A crontab job was installed at every machine to update the indexes.

The main purpose of the second test was to find any bugs in Rumour, and to see if the file structure of Rumour needed to be modified. It turned out that some Web servers do not use a partial URL system (URLs with tilde symbol indicating the user's home directory). Therefore, Jack had to be modified so it could make a Lindex that does not use partial URLs. Other problems encountered with Rumour during these tests included :

• Collector had encountered error files instead of an error message when trying to load URLs that did not exist, therefore collector had to be modified so it could recognise the Cindex files.
• Some Web servers do not follow symbolic links from the cgi-bin directories, therefore search.pl had to be copied into the cgi-bin directory.
  

No real test can be carried out of the effectiveness of Rumour until it is widely installed on Web servers because Rumour is a distributed system. But in general, Rumour is almost ready to be distributed.