You're right that it is the LRM character, and so shouldn't be stripped in general (so it's sensible that it doesn't match \s). But it is useless at the end of a string, hence my suggestion that it should be considered something like whitespace in that context. I hoped there might be a function to trim the end of strings for this specific purpose. Or if not, something generic I could add to a RE to strip unicode characters of this nature.

Give URI::Encode a try.

#!usr/bin/perl -l

use strict;
use warnings;
use URI::Encode qw(uri_decode);

my $encoded = 'http://commons.wikimedia.org
  /wiki/File:Atelerix_algirus.jpg%E2%80%8E';
print uri_decode($encoded);
[download]

Yes, I've done that. It converts the %E2%80%8E string to the unicode LRM character, which isn't printed, but is still embedded in the string, causing problems when accessing the URL again. Thanks, though.