Personally, I would say that you are using the wrong data structures. The keys of you primary hash are 'pc' + n. The pc part is repetative and reduntant. Your left with n, which gives you an array.

In you secondary hashes, your keys are all identical and only the values vary. The greatest benefit of a hash is the ability to look things up by name, but your not able to use this effectively. All your keys are again, repetative and redundant.

I think I would use something like this.

#! perl -slw
use strict;

my( %users, @linux, @win32 );
my @pcs = (
    [ 'win32', 'dave' ],
    [ 'linux', 'john' ],
    [ 'win32', 'robin' ],
);

for( 0 .. $#pcs ) {
    my $pc = $pcs[ $_ ];
    $users{ $pc->[1] } = $_;
    push @{ $pc->[0] eq 'linux' ? \@linux : \@win32 }, $_;
}

print 'Linux users: ', join ' '
    , map{ $_->[1] } @pcs[ @linux ];
print "Dave's pc is number $users{ dave } attributes:"
    , join ' ', @{ $pcs[ $users{ dave } ] };
print "Win32 pc numbers: @win32\n\t"
    , join "\n\t", map{ join ':', @$_ } @pcs[ @win32 ];

__END__
Linux users: john
Dave's pc is number 0 attributes:win32 dave
Win32 pc numbers: 0 2
        win32:dave
        win32:robin
[download]

This scheme allows you to answer all of the questions you mentioned using direct or 1 level indirected indexing with no searching.