Readme-fakewww

This is a brief description of the fakewww server program.

This code was originally developed by Tom Dunnigan (ORNL) to
allow testing of a web based web100 speed tool.  It has been
modified by Rich Carlson (Internet2) to perform several additional
tasks.

The basic operation is to start the 'fakewww' process on the 
server.  This causes the process to start listening for connections 
on port 7123 (use the -p option to specify a different port number).

Once in the listen state, it monitors the connection for new clients.
An incoming connection causes a child process to be fork'ed off while
the parent return to the wait state.  The child process then waits
for an HTTP 'GET' command to arrive.  If anything other than a 'GET'
is recieved, the connection is dropped.

The child then processes the 'GET' command.  It has a compilied in
list of html and java class/jar files that it can return to the client.
If the user specifies anything other than these compiled in files, a
file not found message is returned and the connection is terminated.

By default the fakewww program is installed in the /usr/local/sbin
directory.

There are 6 run time options that may be specified on the command line
     $ /usr/local/sbin/fakewww -h
     Usage: /usr/local/sbin/fakewww {options}
     	     -d 	increment debugging mode
	     -h 	print this help message
	     -l log_FN 	log connection requests in specified file
	     -p port# 	specify the port number (default is 7123)
	     -t # 	specify maximum number of hops in path (default is 10)
	     -F 	operate in Federated mode

The '-F' flag controls the basic operating mode of the fakewww
program.  If the -F is not specified, then the server acts in stand-
alone mode, just as previous versions of the tool did.  If the -F
options is specified, then the server operates in federated mode.

In federated mode, multiple servers work together to form a single
testing domain.  In this mode, the clients are outside the testing
domain and each fakewww server will re-direct the client to the 
'closest' server, i.e., the testing domain ingress server.  This allows
testing of the shortest network path, and any host based configuration
problems will be detected and reported.  Future versions will allow
the user to test to the egress point for better performance diagnostics.

See the readme-FM file for details on setting up a federation of NDT
servers.

Start the fakewww server with the -F options "/usr/local/sbin/fakewww -F".
This will cause the server to perform these additional tasks.

When a new client connects, the server will perform a traceroute back
to the client.  (The number of hops that must be crossed is set by the
-t flag on the fakewww server.)  This task is performed by the fakewww
child process using a modified version of traceroute.  (note this means
that the fakewww process must run as root in order to process the returned
ICMP messages.)

After the traceroute to the client is complete, the fakewww child will
compare this route to that pre-stored map data.  A match is found when
the paths start to diverge.  This ensures that the client will always
recieve a positive answer, but that it might re-direct to itself.  A
dynamic web page is created and sent to the client, indicating that 
the 'closest' NDT server is being contacted and testing can begin once
the java applet is downloaded.