Yes, it's generally best to avoid unnecessary eval STRING, so a simple iteration over the result of spliting the string should do it e.g

my $h = {
  'misc' => {},
  'docs' => {
    'howtos' => {
      'email' => {
        'index.html' => {
          'date' => '21-1-2004',
          'size' => '691'
        }
      },
      'ftp' => {},
      'ssh' => {}
    }
  }
};

use Data::Dumper;

print Dumper( find_hash($h,"/docs/howtos/email/") );

sub find_hash {
  my($tree, $path) = @_;

  my $n = $tree;
  for(grep length, split '/', $path) {
    return unless exists $n->{$_} and defined $n->{$_};
    $n = $n->{$_};
  }

  return $n;
}

__output__

$VAR1 = {
          'index.html' => {
                            'date' => '21-1-2004',
                            'size' => '691'
                          }
        };
[download]

That could probably do with a bit more data validation but it should illustrate, roughly, a saner approach to iterating through your nested hash.