Astonishingly, you are wrong about the character class. [^\w\s] matches characters that are neither words nor whitespace; i.e., punctuation:

$_='o.ne<a>tw/o<b>';
@punks = /[^\w\s]/g;
print "<@punks>";
[download]

yields

<. < > / < >>
[download]

The problem with my example was that I left the /g modifier off the pattern match. I've updated it, and tested it:

while(<DATA>){
/http:\S*[^\w\s]/g and s/\G\n//;
print;
}
__DATA__
there is an http://whatever.com/address/
crossing/line/boundaries.html right in the
middle of this nice string.
[download]