What you have would work if your pipe had sort and uniq in the other order:

perl -ne ... | sort | uniq
[download]

For that matter, you don't need uniq as a separate process:

perl -ne ... | sort -u
[download]

and of course, your perl one-line could do everything:

perl -ne 's/\W+/ /g; for $w (split){$u{$w}++} END{print join $/,sort k
+eys %u,""}'
[download]

For that matter, you could also modify that slightly to print uniq words with their frequencies of occurrence (by including the values of %u along with the keys), which is also handy.