Attributes are like post-it notes you can tack onto anything in Perl. The technical term is "out of band data" I believe.

Attributes are used extensively in Catalyst to declare the URLs to which a subroutine is bound, and I've used them to declare access control lists together with the subroutine. They work by giving you a callback when declaring the subroutine, so they are more or less syntactic sugar for:

Ultracool->declare_subroutine('frobnicate',sub {
  ...
}, qw(some more arguments));
[download]

Wow - that's really the exact same question, only better phrased :-)

Thanks! Interesting discussion. I think you might use them for example to distinguish between types of methods, e.g.:

sub set_name : Interface {
    croak "Not implemented!";
}
sub get_name : Interface {
    croak "Not implemented!";
}
[download]

So that you can query whether a method call you are about to make is implemented by the right class - and not by SUPER:: in an interface - without using eval { ... } blocks.

I haven't seen it mentioned yet, but this is documented in perlsub. See the "Subroutine Attributes" section for all the gory details.

Update: oops, I just noticed from the OP:

I've read perldoc perlsub and perldoc attributes

Apologies for not reading the entire post.