GRADER GUIDE TO MOSS

INTRODUCTION - Moss(Measure Of Software Similarity) is a system used for checking for similarity in a large
               number of programs. Moss is created and maintained by people at Berkeley. Here is the official
               Moss site: http://www.cs.berkeley.edu/~aiken/moss.html.

GETTING      - First you should visit the Moss script download site here: http://www.cs.berkeley.edu/~aiken/moss.html.
STARTED        I suggest downloading the 1.2 script for Unix. Next you'll need a Moss account. For this you
               should email moss-request@cs.berkeley.edu. Explain in the email that you are a grader from UCR.

               Now wait for about a day. You should get back a reply mail that contains a Moss script. If you'd
               like, save this email and delete all the headers making sure the #!/usr/bin/perl is the first
               line in the file. Or if you want to use the script you downloaded earlier then scroll down until
               you see the $userid=784784734(some number). Make a note of this number, it's your login id. Now
               edit the script you downloaded, and find the $userid=987654321 line, and replace the bogus ID with
               your own.

               One last thing, you'll need to make the script executable. Do a "chmod 700 filename", where filename
               is the name of the script.

USING MOSS   - The commented out portion of the MOSS script explains all the parameters, so if you want to do
               something strange then see the documentation. I'll just go over a sample use.

               Suppose your directory structure is set up so that in your home you have a directory called TURNIN. 
               Inside TURNIN are the directories for the individual assignments, AS1-AS9 or whatever. Inside each of 
               these are the student directories, which contain a bunch of turnin directories and a symlink called ONTIME 
               that points to the latest. You probably are familiar with the turnin system already. Suppose you wanted to
               cheat check all the AS3 c++ files, and the name of the script is MOSS. You'd do this:

               $ MOSS -l cc -d ~/TURNIN/AS3/*/ONTIME/*.{cc,cpp}

               The "-l cc" paramater tells Moss that you are submitting c++ files. The -d tells Moss to compare by 
               directory. You final parameter are the files that you are submitting, every AS3 ONTIME .cc and .cpp
               file. You screen should now be flooded with "file submitted" messages, and finally a note telling
               you to wait a while and check your mail later for the results. Go to the bathroom for ten minutes
               and you should see a return mail from Moss. It will have a URL for you to visit. Follow the URL
               and you'll see the results of your submission.

               The interface is pretty intuitive. You'll see a list of similar assignments sorted by lines matched.
               You'll also see a similarity percentage. More on this later. You can click on a file name to have it
               lined up with the one it matches. You can click the bullets to scroll to the similar sections.

JUDGEMENT    - I've heard many people ask "what's a good threshold for determining cheating". There really isn't one.
               What you should do is take a look at the outliers. If hundreds of pairs have 10% similarity, and then
               there are 15 that have 80%+ then take a look at those 15 and see if they are too similar. Moss is a
               tool that can help catch cheaters, but it should not be used without human judgement.


               
               If something is wrong or missing here then send a mail to snematbakhsh@cs.ucr.edu.