Do whatever fits your needs.

I prefer to have short scripts. If the problem is small and special enough to be solved by a few routines stuffed into a single script file, I usually do exactly that.

For larger problems, I prefer spreading my code over separate classes / packages / files, with a quite minimal loader script. Typically, I create a new (sub-)project in subversion, i.e. a directory containing at least the three subdirectories bin, etc, and lib. The loader script goes into bin, the perl packages into lib, and some kind of configuration file into etc. The loader script typically looks like this:

#!/usr/bin/perl -T

use 5.010;
use strict;
use warnings;

use lib '/some/hardcoded/path/lib';
# or:
#use FindBin;
#use lib "$FindBin::Bin/../lib";

use Some::Prefix::MainClass;

Some::Prefix::MainClass->main(@ARGV);
[download]

Sometimes, there are several similar loader scripts for slightly different purposes, e.g. for programs that can be used both on the command line and via CGI / FastCGI.

Sometimes, a set of classes becomes so useful that they end being installed as regular Perl modules. But most Perl modules I've written linger around in project specific lib directories and are only available to programs that manipulate their @INC, i.e. loader scripts.

Alexander