FusionInventory Agent
			---------------------

Dependencies
------------

Minimum perl version: 5.8

Additional mandatory perl modules:
- Digest::MD5
- XML::TreePP
- LWP
- Net::IP
- UNIVERSAL::require
- Mac::SysProfile (MacOs only)

Additional optional perl modules:
- Compress::Zlib, for message compression
- HTTP::Daemon, for web interface
- IO::Socket::SSL, for HTTPS support with certificate validation
- Net::CUPS (v0.60 minimum), for printers detection 
- Proc::Daemon, for daemon mode (Unix only)
- Proc::PID::File, for daemon mode (Unix only)

Additional optional programs:
- dmidecode, for hardware inventory
- lspci, for hardware inventory

SSL support
-----------
LWP (also known as libwww-perl) is able to use either Crypt::SSLeay or
IO::Socket::SSL transparently as underlying socket implementation for HTTPS
support. However, only the second one is able to perform server certificate
validation. As a consequence, the agent will refuse to use HTTPS, and exit
immediatly if IO::Socket::SSL is not available, unless certificate checking has
been explicitely disabled (--no-ssl-check command line parameter, or
no-ssl-check configuration parameter).

System-specific informations
----------------------------

On Solaris, Sun Studio seems to be needed to build the dependency. The
generated Makefile needs gmake to be executed. The default installation will
install the binary in /usr/perl5/5.XXXXX/bin, set your $PATH variable according
to that.

On Solaris/SPARC, you must install sneep and record the Serial Number with it.
Download it from http://www.sun.com/download/products.xml?id=4304155a.

On Windows, if you don't want to use the installer for Windows you can follow
the installation process avalailable from
http://forge.fusioninventory.org/projects/fusioninventory-agent/wiki/RunAgentWin32.

Usage
-----
They are two different ways to run the agent:
- as a temporary process, exiting after completing its run
- as a permanent process, staying in memory permanently (daemon under Unix,
  service under Windows)

In the first case, the agent execution scheduling is completly controlled by
the agent, through a dedicated system, such as cron under Unix, or XXX under
Windows. In second case, the agent execution scheduling is server-controlled,
as the agent runs regulary according to a parameter initially set in its
configuration (delaytime), overrided by the server at the first connection
(PROLOG_FREQ). Additionaly, the server may also initiate additional
out-of-schedule executions by sending HTTP requests if the web interface is
used.

When executing, the agent tries to runs every available task for every
configured target. A task is a specific kind of work to perform (local
inventory, network discovery, wake on lan, etc...), while a target is both a
recipient for the result and a controller for this work (an OCS server, a GLPI
server, a local directory, etc...). Only the local inventory task is compatible
with all kind of targets, tough, all others require a GLPI server target, and
will get discarded for other kind of target.