The point is control. Eval-based dispatch is much like an unstructured goto $EXPR - you cannot really be sure of the program flow. An explicit hash lists all locations where the code is allowed to branch to. This prevents malicious attacks where the attacker could try to make a call to the exit or system core function and it allows much better error handling. It also decouples the subroutine name from the argument, which I find very convenient in web applications.

dominus wrote a series of articles on this topic that explains the problem very nicely. Definitely worth the read.

They think it's a bad idea because 'use strict' complains ... but if you know that what you're trying to do is correct, you can shut off 'use strict' for the specific complaint that it has:

use strict;
use warnings;

for my $i ( 1 .. 2 ) {
  $a = 'test'.$i;
  no strict 'refs';
  &$a;
}

sub test1 { print 'test1'; }
sub test2 { print 'test2'; }
[download]

So where's the problem with that? None, really. I mean, it doesn't propogate, so you're still protected in the subroutine you're calling:

use strict;
use warnings;

for my $a qw( test1 test2 ) {
  no strict 'refs';
  &$a;
}

sub test1 { print 'test1'; }
sub test2 { my $x='test3'; &$x }
[download]

PS. For the original poster -- it's convention that you don't use $a and $b as variable names, unless you're dealing with sorting.

You don't actually need to change the current strictures:

use strict;
sub foo {
        print "Foo";
}

sub bar {
        print "bar";
}

main->can("foo")->();
[download]