There is a major con, though. If you plan on using this regular expression in a loop, it will need to be recompiled every time it is used. With qr//, you will only compile once... You may want to benchmark them to find out the difference:

#!/usr/bin/perl -w

use Benchmark;

my $txt = 'find this regex';
my $str = 'regex?';
my $rgx = qr/$str/i;

timethese
  (
   1000000,
   {
    match => sub{ $txt =~ m/$str/i },
    qreg  => sub{ $txt =~ $rgx },
   },
  );
[download]

Gives me the following:

% test.pl
Benchmark: timing 1000000 iterations of match, qreg...
     match:  1 wallclock secs ( 0.82 usr +  0.00 sys =  0.82 CPU) @ 12
+19512.20/s (n=1000000)
      qreg:  0 wallclock secs ( 0.73 usr +  0.00 sys =  0.73 CPU) @ 13
+69863.01/s (n=1000000)
[download]

As you can see, for a regex this simple, it doesn't matter too much... but more complicated regexs wll show a bigger difference.