...because greediness does not affect its result in any way.

I was hoping someone would have an idea as to what he might have been thinking.

You're only thinking about the final outcome. Sometimes, changes are not made for the outcome, but because it changes how it performs, and one method may be better for the normally expected input.

Unfortunately, in this case, I'm not seeing that on my system the non-greedy being significantly better at matching the short version, and it's significantly worse at matching the long version:

use Benchmark qw(cmpthese);

my $f  = '-f';
my $f2 = '-field';

cmpthese( 500000, {
    'ng_short' => sub { $f  =~ /^-f(i(e(ld??)??)??)??$/  },
    'ng_long'  => sub { $f2 =~ /^-f(i(e(ld??)??)??)??$/  },
    'g_short'  => sub { $f  =~ /^-f(i(e(ld?)?)?)?$/  },
    'g_long'   => sub { $f2 =~ /^-f(i(e(ld?)?)?)?$/  },
    }
);

#              Rate  ng_long   g_long  g_short ng_short
# ng_long   88339/s       --     -54%     -73%     -74%
# g_long   193798/s     119%       --     -41%     -42%
# g_short  328947/s     272%      70%       --      -2%
# ng_short 335570/s     280%      73%       2%       --
[download]

Oh -- and it _does_ change the values of $1, $2, $3, but we have no reason from the code snippet given that they're important.,/strike>