Again, just use LWP (and possibly Time::HiRes if you want finer than second granularity).

use LWP::Simple qw( get );
use Time::HiRes qw( time );

my $url = "http://...";

my $start = time;
getstore( $url, "/dev/null" );
my $end = time;

print "Took ", $end - $start, " seconds\n";
[download]

What's wrong with

use LWP::Simple qw( get );
use Time::HiRes qw( time );

my $url = "http://www.perlmonks.org/";

my $start_time = time;
get($url);
my $end_time = time;

print("The transfer took ", $end_time-$start_time, " seconds.\n");
[download]

You could easily put that into a loop and take the average (but beware of caching proxies).

Update: oops, I was a bit slow at posting this!

My weblint program, available on CPAN, has a download timer:

$ weblint++ -t http://www.perlmonks.com
2.315 seconds
[download]

It's probably easier to do it yourself as some of the other posts showed you, though.