I imagine it could be possible for a server to go through the motions of downloading the file and then forward the results (during the HTTP exchange in the front-end) to a user, but it kind of defeats the purpose of using a distributed network in the first place.

For the network to function as designed, each user would need to be running a gnutella engine of sorts.