Dude.. you guys don't understand how the network work at all. If you jag up the connection number as a leaf node. You get a bigger footprint on the network by displacing other leaf nodes. So, you ended up with more search results yourself by reducing the total network size.

For example, lets say the total network size is 8 and you can only search result 1 up and 1 down. So, if you come in as slot 4, you see results from 3 and 5. Search results come from 2 nodes out of 8 unique leaf node. Let's say if you are unsatified with the result and jag up your connection to 2. Now you occupied slot 4 and 6 now. Search results come from 4 nodes out of 7. You increase your search results by shrinking the total network size. Let's say if you are still unsatified with the result and jag up your connection to 8. Guess what, you are the network and you can get search results by just opening your harddrive.

Of course, this assumption don't hold true if the network is fragmented. This assumption also don't hold true if there are more ultrapeers nodes than leaf nodes. (Since there are people complain about connections can't stay up, I think this is not true anymore)

Anyway, I have to agree with limewire here.

To get good search results as a leaf node, you have to hook up with a good ultrapeer node (One that has a lot of leaf nodes). That means a ultrapeer node that stay around for a long time. You can't tell which one is good unless you-- yourself stay up for a while and let's the bad ultrapeer node switch out themselves.

In terms of search results, gnucleus seems to do a better job than limewire but they use up more bandwidth.