To explain the first response syntax:

#declarations
my ($quote_on,$output);

#default scalar is the data
$_ = q/this is a "test of some regexp" blah blah
this is another "test of some regexp" foo bar/;

#using $_ begin a for loop performing a split on each #element

for (split //) {

#quote_on is a boolean set to true when the increments reach
#a quote (and then off again when reaching another quote)

 $quote_on = ($quote_on ? 0 : 1) if (/\"/);

#if processing elements after a quote, then replace the
#whitespace with an underscore

 s/\s/_/g if $quote_on;

#assign all that to $output

  $output .= $_;
}
[download]

I didn't write the original post, but I was thinking about doing it a different way, although this one seems to be
perfectly adequate and more importantly, maintainable, than a seemingly more elegant one-liner. Just IMHO