It's the reason why you cannot use "my" variables in the top level of source files for mod_perl, when using Apache::Registry, the easy route from CGI to mod_perl.

There are articles on how to fix variables for inner subroutines, but the fact that you even need to worry about it is a major counterpoint against Perl, in my opinion.

The reason, as far as I gather, is because BEGIN (and related) blocks are actually subs, that are executed once; you can even use the syntax

sub BEGIN {
    ...
}
[download]

And, Larry Wall and other prominent figures have said, years ago, that it'll never be "fixed". Because it is so by design.

It's got nothing to do with BEGIN being sub-like. Inner subs would have to capture every time the outer sub is called. But since there are no private subs in Perl, there isn't necessarily any executing outer sub to capture from.

sub outer {
   sub inner {
   }
}
[download]

is basically

sub outer {
   BEGIN {
      *inner = sub {
      };
   }
}
[download]

Replace it with

sub outer {
   *inner = sub {
   };
}
[download]

and inner will capture properly for an inner sub. Of course, problems will still ensue if you call inner() from outside of outer(). To solve that, you'd need a language feature such as

sub outer {
   my sub inner {
   }
}
[download]