A nice way to do this is the way provided by the LWP::UserAgent module. It allows you to provide a code reference that gets invoked whenever it gets another chunk of a file. Check this:

my $sub = sub {
                my ($data, $res, $proto) = @_;
                
                # either write directly to file or push onto an array
                $filename ? print FILE $data : push(@res_lines, $data)
+;
                
                # increment total length
                $data_recved += length $data;
                
                if ($already_drew_percentage) {                
                    # erase an old sign and draw a new one
                    
                    print "\b" x 9, sprintf('%03d%% done',
                         (($data_recved / $res->content_length) * 100)
+);
                } else {
                    # draw the first percentage
                    
                    printf('%03d%% done',
                         (($data_recved / $res->content_length) * 100)
+);
                   $already_drew_percentage = 1;
           }
};
[download]

(Sorry about the tab mess...I'm not sure how to correct it.) Call it like this:

Where $agent is an LWP::UserAgent, $req is an HTTP::Request and 1024 is the amount of bytes you want to get before invoking $sub.

That was ripped from the pronbot source. It allows you to either write binaries directly to file via the FILE handle and the value of $filename, or push the lines of the response onto an array which can later be joined. (You might have do declare some of the vars yourself :) I'd advocate using an LWP module for basic HTTP stuff, it's very easy, although this probably looks complex.