Well I was almost ready to post a $ENV{SCRIPT_FILENAME} option, but before I did I ran a benchmark on the three options (includes the option of the wrong filename). It seems that the Win32::GetLongPathName is in deed the fastest with the correct filename. Here is my simple Benchmark code:

#!perl
use strict;
use File::Basename;
use CGI;
use Benchmark;
use Win32;

my $cgi = CGI->new();

print $cgi->header;

print "Hello World<br>";

print "DOS NAME: " . basename($0) . "<BR>";

print "ENV BASENAME: " . basename($ENV{SCRIPT_FILENAME}) . "<br>";

print "Win32 BASENAME: " . basename(Win32::GetLongPathName($0)) . "<br
+>";

timethese (100000, {
    'basename_env'   => sub { basename(Win32::GetLongPathName($0)) },
    'basename_win32' => sub { basename($ENV{SCRIPT_FILENAME}) },
    'basemane_dollar0' => sub { basename($0) },
    }
    );
1;
[download]

The results

Benchmark: timing 100000 iterations of 
    basemane_dollar0, 
    basename_env, 
    basename_win32...
basemane_dollar0:  5 wallclock secs 
    ( 3.68 usr +  0.00 sys =  3.68 CPU) @ 27203.48/s (n=100000)
basename_env: 121 wallclock secs 
    (34.99 usr + 81.07 sys = 116.06 CPU) @ 861.65/s (n=100000)
basename_win32:  5 wallclock secs 
    ( 4.24 usr +  0.00 sys =  4.24 CPU) @ 23557.13/s (n=100000)

Now I would like to know why the $ENV way is SO slow. I had expected there might be a little difference, but 10 times slower?

UPDATE: The above code has a glaring error. The subs for env and win32 were flip flopped in the timethese function. The Win32 way is 10 times slower not the ENV way. I apologize for the late night code :^)