The time I wrote such a wizard, I followed the following criteria :

• All data is collected within the wizard until the last page via hidden fields. This took away the burden of server-side session management and allowed for easy automation, as any program would simply have to feed the last page of the wizard with all the data in one request.
• All pages had names (instead of numbers), which I then correlated to the functions that displayed the page via a hash. This allowed me to nicely group the interesting stuff of one page together in one place in the code.
• The navigation had to be via submit buttons, as otherwise the passing of hidden fields wouldn't have worked. This of course broke the navigational features of the browsers, but in a way that didn't create maintenance or diagnosis headaches.

The resulting wizard framework was nice, as adding another page meant adding another sub, connecting that sub to the name of the page and inserting the name of the page in the sequence of pages. Adding more fields to the wizard also only meant to make the new variables known to the wizard (so that they would survive changing the page) and adding a way to set those values.

perl -MHTTP::Daemon -MHTTP::Response -MLWP::Simple -e ' ;    # The  
$d = new HTTP::Daemon and fork and getprint $d->url and exit;#spider
($c = $d->accept())->get_request(); $c->send_response( new   #in the
HTTP::Response(200,$_,$_,qq(Just another Perl hacker\n))); ' #  web
[download]