Hi,

Step one to fixing this is, forget the program exists, and define your goals

For example , mirror the title/alt/image of all xkcd , so that would be

- get xkcd page
- extract info ( id title alt text image next )
- save files 
- repeat with next

Next up tweak the goals a bit, be nice

- get page if not already exist
- extract info ( id title alt text image next ) and de-html-textify
- save files with safe filenames 
- repeat with next
- wait andor quit, when done quit, when limit reached wait or quit until next time, keep track of progress

Next is write (code) the program of goals

save_xkcd( 'outdir', 'startingid' );
sub save_xkcd {
    $starting_id ||= id_from_progress();
    my @ids = $starting_id;
    while( @ids ) {
       my $cid = shift @ids;
       my $page = sptintf '...%s', $cid;
       $mech->get( $page );
       save_stuff( $mech, $cid );
       next_page( $mech , \@ids );
       maybe_sleep();
    }
}
[download]

Now all you do is fill in the blanks

No need for CGI in this equation, cgi doesnt like near infinite loops anyway

$mech->title gets you de-htmld text like   xkcd: House of Pancakes

HTML::TreeBuilder::XPath gets you the alt/title text with xpath query of '//img/@title' and next link with a query of '//a[@rel="next"]'

Or  $mech->find_link( text_regex => qr/next/i );

Yes, you could fix up your program by replacing LWP::Simple with mech ... but thats not exactly fun now isnt it :)