because that would fork-bomb a machine if the array is large.

i did this in a previous script:

my $fork_num = ( $#hostlist <= MAXFORK ? $#hostlist : MAXFORK );

for ( 0..$fork_num ) {

    ##
    ## create bi-directional socket for parent/child communication

    my $to_child = IO::Socket->new();
    my $to_parent = IO::Socket->new();

    socketpair($to_child, $to_parent, PF_UNIX, SOCK_STREAM, PF_UNSPEC)
        or die "Couldn't create socket pair:  $!\n";

    my $pid;
    if ( $pid = fork ) {
                
        ### parent process

        undef $to_parent;
        $to_child->autoflush;

        $children{$pid} = $to_child;

        sleep 1;
        next;

        }

    else {

        ### child process

        undef $to_child; undef %children;
        $to_parent->autoflush(1);

        local $SIG{PIPE} = sub { print " CAUGHT PIPE\n\n";
                                                die "Signal caught in 
+Child $$:  $!\n"; };

        ## 
        ## main processing occurs in the subroutine.  

        &process_config_job($fork_num, $to_parent);

    }

}
[download]

the child processes have a sub getnexthost() which asks the parent for another hostname.

the parent process has a subroutine to read the request for a new hostname then send it.

i don't like the 'fork foreach' idiom. good way to crash a box.