DBI::ProxyServer - a server for the DBD::Proxy driver
SYNOPSIS
use DBI::ProxyServer;
DBI::ProxyServer::main(@ARGV);
DESCRIPTION
DBI::Proxy Server is a module for implementing a proxy for the DBI proxy
driver, DBD::Proxy. It allows access to databases over the network if the
DBMS does not offer networked operations. But the proxy server might be
usefull for you, even if you have a DBMS with integrated network
functionality: It can be used as a DBI proxy in a firewalled environment.
DBI::ProxyServer runs as a daemon on the machine with the DBMS or on the
firewall. The client connects to the agent using the DBI driver DBD::Proxy,
thus in the exactly same way than using DBD::mysql, DBD::mSQL or any other
DBI driver.
The agent is implemented as a RPC::PlServer application. Thus you have
access to all the possibilities of this module, in particular encryption
and a similar configuration file. DBI::ProxyServer adds the possibility of
query restrictions: You can define a set of queries that a client may
execute and restrict access to those. (Requires a DBI driver that supports
parameter binding.) See ``CONFIGURATION FILE''.
OPTIONS
When calling the DBI::ProxyServer::main() function, you supply an
array of options. (@ARGV, the array of command line options is used,
if you don't.) These options are parsed by the Getopt::Long module.
The ProxyServer inherits all of RPC::PlServer's and hence Net::Daemon's
options and option handling, in particular the ability to read
options from either the command line or a config file. See
RPC::PlServer(3). See Net::Daemon(3). Available options include
chroot (--chroot=dir)
(UNIX only) After doing a bind(), change root directory to the given
directory by doing a chroot(). This is usefull for security, but it
restricts the environment a lot. For example, you need to load DBI
drivers in the config file or you have to create hard links to Unix
sockets, if your drivers are using them. For example, with MySQL, a
config file might contain the following lines:
my $rootdir = '/var/dbiproxy';
my $unixsockdir = '/tmp';
my $unixsockfile = 'mysql.sock';
foreach $dir ($rootdir, "$rootdir$unixsockdir") {
mkdir 0755, $dir;
}
link("$unixsockdir/$unixsockfile",
"$rootdir$unixsockdir/$unixsockfile");
require DBD::mysql;
{
'chroot' => $rootdir,
...
}
If you don't know chroot(), think of an FTP server where you can see a
certain directory tree only after logging in. See also the --group and
--user options.
clients
An array ref with a list of clients. Clients are hash refs, the attributes
accept (0 for denying access and 1 for permitting) and mask, a Perl
regular expression for the clients IP number or its host name. See
``Access control'' below.
configfile (--configfile=file)
Config files are assumed to return a single hash ref that overrides the
arguments of the new method. However, command line arguments in turn take
precedence over the config file. See the ``CONFIGURATION FILE'' section
below for details on the config file.
debug (--debug)
Turn debugging mode on. Mainly this asserts that logging messages of
level ``debug'' are created.
facility (--facility=mode)
(UNIX only) Facility to use for ``Sys::Syslog (3)''. The default is
daemon.
group (--group=gid)
After doing a bind(), change the real and effective GID to the given.
This is usefull, if you want your server to bind to a privileged port
(<1024), but don't want the server to execute as root. See also
the --user option.
GID's can be passed as group names or numeric values.
localaddr (--localaddr=ip)
By default a daemon is listening to any IP number that a machine
has. This attribute allows to restrict the server to the given
IP number.
localport (--localport=port)
This attribute sets the port on which the daemon is listening. It
must be given somehow, as there's no default.
logfile (--logfile=file)
Be default logging messages will be written to the syslog (Unix) or
to the event log (Windows NT). On other operating systems you need to
specify a log file. The special value ``STDERR'' forces logging to
stderr. See Net::Daemon::Log(3) for details.
mode (--mode=modename)
The server can run in three different modes, depending on the environment.
If you are running Perl 5.005 and did compile it for threads, then the
server will create a new thread for each connection. The thread will
execute the server's Run() method and then terminate. This mode is the
default, you can force it with ``--mode=threads''.
If threads are not available, but you have a working fork(), then the
server will behave similar by creating a new process for each connection.
This mode will be used automatically in the absence of threads or if
you use the ``--mode=fork'' option.
Finally there's a single-connection mode: If the server has accepted a
connection, he will enter the Run() method. No other connections are
accepted until the Run() method returns (if the client disconnects).
This operation mode is usefull if you have neither threads nor fork(),
for example on the Macintosh. For debugging purposes you can force this
mode with ``--mode=single''.
pidfile (--pidfile=file)
(UNIX only) If this option is present, a PID file will be created at the
given location. Default is do not create a pidfile.
user (--user=uid)
After doing a bind(), change the real and effective UID to the given.
This is usefull, if you want your server to bind to a privileged port
(<1024), but don't want the server to execute as root. See also
the --group and the --chroot options.
UID's can be passed as group names or numeric values.
version (--version)
Supresses startup of the server; instead the version string will
be printed and the program exits immediately.
CONFIGURATION FILE
The configuration file is just that of RPC::PlServer or Net::Daemon
with some additional attributes in the client list.
The config file is a Perl script. At the top of the file you may include
arbitraty Perl source, for example load drivers at the start (usefull
to enhance performance), prepare a chroot environment and so on.
The important thing is that you finally return a hash ref of option
name/value pairs. The possible options are listed above.
All possibilities of Net::Daemon and RPC::PlServer apply, in particular
Host and/or User dependent access control
Host and/or User dependent encryption
Changing UID and/or GID after binding to the port
Running in a chroot() environment
Additionally the server offers you query restrictions. Suggest the
following client list:
then only the users root and wwwrun may connect from admin.company.com,
executing arbitrary queries, but only wwwrun may connect from other
hosts and is restricted to
$sth->prepare("select");
or
$sth->prepare("insert");
which in fact are ``SELECT * FROM foo'' or ``INSERT INTO foo VALUES (?, ?, ?)''.
The DBI::ProxyServer module is free software; you can redistribute it
and/or modify it under the same terms as Perl itself. In particular
permission is granted to Tim Bunce for distributing this as a part of
the DBI.