If you are trying to spider an FTP site, you would be better off just using plain wget or a purpose-built multi-threaded spidering tool like Heritrix. The network overhead of setting up new connections would likely dwarf any benefit from parallelism, especially in the case of many small files, the only case you stand a chance of being cpu- rather than bandwidth-limited anyway -- to say nothing of avoiding attempts to download files that don't exist. Keep it simple.