The Gnutella network came into being because when a commercial project to develop file sharing was being tested across the internet, it was copied by some enterprising people. They reverse-engineered the code to form what you now see. They had a choice as to the direction in which to release it. The original intention had been to allow file sharing and chat between designated machines — to make LANs and WANs into a more effective community. This was felt too narrow so the current open-ended daisy-chain version was released. This simply connects all machines switched on at any given time and allows connections to be established between them almost at random (through the chance of individual machines being available or not).

So the spirit of what you are asking was the designers' original intention: that you would link specific machines together through their addresses. And that is why browsing and chat works (or, in your case, does not but, frankly, chat is not particularly useful since little of value can be said to a random connection in a short space of time). Every machine on the net is allocated a unique address by the ISP (Internet Service Provider). This is broadcast everytime you connect. But Gnutella's present configuration does not support direct targetting. In a different context, you would be asking how to hack into particular machines, the only difference here being that everyone who shares files on Gnutella is actually inviting others to take that group of files.

The developers continuously discuss how Gnutella should be "improved" (it is all very democratic). Obviously, one possibility would be to allow stable connections to be created between specific machines for chat and file exchange purposes. Depending on how quickly agreement is reached, you are likely to see new features appearing in the Gnutella access packages like LW, for example to allow one site to create a magnet for other users to link for specific download purposes. LW is a "work-in-progress. It is evolving to meet the needs and expectations of its users but, because it is not a commercial venture, the actual directions taken may be more communal than specific. It will be interesting to see.

The search results you achieve are saved in a temporary file which is deleted at the end of each session. If your understanding of your own machine is good enough, you can copy this file before quitting LW but it will not help you because LW does not currently allow you to input those addresses for search purposes. However, if you do not specifically delete partial downloads from the download box, they are reinstated when you start your next LW session and LW will automatically attempt to reconnect to those original hosts. Sometimes, this does reinstate the connection if, by chance, the host is available. Similarly, refreshing the original searches can also reconnect you. When the hosts can be browsed, you can also browse from the download box. The choice is yours. If you keep large numbers of partial downloads in this way, you keep the chances of completing those downloads alive. But the greater the number of these chances, the more slowly your package will work. Personally, I delete everything that fails and simply (re)search for the particular files over time when I reconnect. Don't forget that, sometimes, only one or two hosts have what you want and once you have exceeded their download quota, you cannot go back for a few days. Keeping lists of what you want is a good idea.

If you are trying to identify particular cultural types of file for downloading, work on your search strategies. There may be language problems if what you want may be stored using different alphabets. Try changing your keyboard configuration to the relevant languages for search purposes and see if that helps.

And keep trying. It is only by experimenting that you will learn.