Thanks. This version runs ok - I tried it on directory containing 77 html files, and it was pretty quick. What quantity of data are you handling? How big is the memory footprint?

Just a couple other suggestions:

• fix your indentation - making the code more legible really helps.
• create the cleaned output html files in a separate directory, leaving the original input files unaltered, so that you can do multiple runs on the same input data, compare input to output, and compare outputs using different setups
• use @ARGV for selecting input and output paths

Something like this:

...
unless ( @ARGV == 2 and -d $ARGV[0] and -d $ARGV[1] ) {
    die "Usage:  $0  input/path  output/path\n"
}
my ( $indir, $outdir ) = @ARGV
my @files = glob "$indir/*.html";
die "No html files found in $indir\n" unless ( @files );
...
for my $file ( @files ) {
    ...
    ( my $ofile = $file ) =~ s{$indir}{$outdir};
    open OUT, '>', $ofile or die "$!";
    ...
}
[download]