*******
NB: you probably don't want to try this with production traffic yet!
*******

Here's how to integrate a mirror server into an ICP based WWW caching
mesh, in four(-ish!) easy steps:

1) edit or create a set of mirror specifications as per the files
     in the 'mirrors' subdirectory:

   * create a separate file for each mirrored site, e.g.
       if you mirrored GNOME you might call the file 'gnome'

   * put all of these specifications in a single directory,
       called say 'mirrors'

   * the first line of the file should be the URL prefix of
       the local mirror, e.g. ftp://ftp.net.lut.ac.uk/gnome

   * the second and subsequent lines should be the URL
       prefixes of the other mirror sites which you would
       like your server to subsume, e.g.
         ftp://ftp://ftp.linux.hr/pub/gnome
         ftp://mirror.aarnet.edu.au/pub/gnome/
       and so on...

2) install a copy of the Squid WWW cache to front end your mirror
     service.  If you don't know what Squid is or where to get it,
     you're probably not ready to play with this stuff (yet).

3) configure Squid to use the starfish.pl program as a redirector
     - depending on where you put your mirror specification files,
     you might need to change the reference to "/usr/local/mirrors"
     at the beginning of starfish.pl

4) run a copy of jellyfish.pl on the same machine (it listens on
     port 3131 only for now), with a command line like so:

   jellyfish.pl /usr/local/mirrors/*

Now, the people in your caching hierarchy should be able to configure
their caches to point at your new Squid server.  Make sure they use
port 3131 for ICP, rather than the normal Squid ICP server (which
runs by default on port 3130).

This is a first release, and the code is still being developed, so
don't be surprised if things either don't work or don't work as you'd
expect them to!

Martin