The first rule of XML::Parser's use: Don't. Or rather, don't use it directly. Unless you really must which is much less often than you might think.

#!/usr/bin/perl -w
use strict;
use XML::Rules;

use LWP::Simple;  # used to fetch the chatterbox ticker

my $cb_ticker = get("http://perlmonks.org/index.pl?node=chatterbox+xml
++ticker");

my $parser = XML::Rules->new(
    stripspaces => 7,
    rules => {
        message => sub {
            my ($tag, $atts) = @_;
            $atts->{'_content'} =~ s/\n//g;

            my ($y,$m,$d,$h,$n,$s) = $atts->{'time'} =~ m/^(\d{4})(\d{
+2})(\d{2})(\d{2})(\d{2})(\d{2})$/;

            # Handles the /me
            $atts->{'_content'} = $atts->{'_content'} =~ s/^\/me// ?
            "$atts->{'author'} $atts->{'_content'}"   :
            "<$atts->{'author'}>: $atts->{'_content'}";
            $atts->{'_content'} = "$h:$n " . $atts->{'_content'};
            print "$atts->{'_content'}\n";
            return;
        },
        'INFO,CHATTER' => 'pass',
    }
);

$parser->parse($cb_ticker);
[download]

Isn't this easier? Now imagine the <message> tag was not so simple, imagine it contained a structure of subtags and subsubtags. Your handlers would have to keep track of where in the structure the parser is and would have to build the datastructure containin that data so that finaly they can access it in the endtag handler if and only iff the tag is <message>. Not what I would call convenient.

With XML::Rules you'd just specify what tags do you want to include (and whether they are supposed to be repeated, contain text content etc. ... the rules may be infered from a DTD or example) and assign a handler specificaly to the <message> tag. And the handler will have access to the datastructure built from the subtags.

With XML::Twig you'll specify the twig_root (or something, I don't remember details) and again will assign a handler to the specific tag and receive all the data from the part of the XML enclosed in it.

And in neither case does the parser have to parse the whole file before your handlers are started and at no time is the whole parsed XML in the memory. (Well, if you use the modules correctly.)