achieving reduced latency in a congested large network

Hello,
    I was thinking to bring this questions to TAG here, and think its
relevant for the greatly distributed nature of www.

I read recently about a network architectural pattern, that its possible
_to achieve reduced long distance latency_ in a congested network like www,
i.e irrespective of the size of network.

My question is, is this merely in theoretical domain, or is physically
achievable by current means?


-- 
Regards,
Mukul Gandhi