You really should take a look at WWW::Mechanize. Mechanize is a subclass of UserAgent that's designed to make just this kind of "walking a site" tasks easy. It also handles cookies by default, so for most sites you shouldn't have any problems handling the session/login information (mechanize should just do the right thing).

Example:

my $m = WWW::Mechanize->new();

# point to the url of the login form
$m->get("http://www.mydomain.com/login_form.html");

# submit form with data specified
$m->submit_form(
   form_name => "login_form",
   {
      username => "xxxxxxx",
      password => "yyyyyyy",
   }
);

# now you are logged in.
# do actions that you need to be logged in for
$m->get("http://www.mydomain.com/program.cgi?something");
# or
$m->follow_link( text => "Click here to perform some action");
[download]

Once you're familiar with WWW::Mechanize, it's a lot easier to automate web applications with it than with LWP::UserAgent, and you can still use (almost) all of the tricks you can do with UserAgent too.

use strict;
use LWP;
my $browser = LWP::UserAgent->new;
  
$browser->credentials(
      'www.mydomain.com:80',
       # Look up this realm with your browser, eg. Firefox
      'web_server_usage_reports', 
      'user_plinky' => 'Password_banjo123'
);

my $url =
  'http://www.mydomain.com/program.cgi?var1=apples&var2=bananas';
 
my $response = $browser->get($url);

#......
[download]

Realize, too, that you can use a module such as HTTP::Recorder or WWW::Mechanize::Shell to record one (or many) successful manual form submission(s). The output of HTTP::Recorder, for instance, can be "dropped" right into your WWW::Mechanize scripts. This makes the development of such scripts trivially easy.

HTH