Question concerning Morpheus
 

Does it really matter the ultrapeers that you conenct to? Going back and looking at Moak's post: http://www.gnutellaforums.com/showth...5&pagenumber=1 -- there are some excellent flash animations of how the gnutella network operates: http://www.milkdrinker.com/GnutellaToadHow4.swf and http://www.limewire.com/gnutella_flash.html

As I see it, esepcailly from reading his posts, the reduced search results are steming from the gnutella network actually becoming too large, and not enough broadband with the majority of newcomers to the gnutella community to prevent searches from timing out. Perhaps a more centralized solution is going to be the best way to go. Or ... perhaps ... seperate client and server versions of the software. Who knows? ... I certianly dont ... just looking for some thoughts on the matter.

It does matter to which ultrapeer you are connected to, but there are a lot of other reasons for having low results. If you had a network neighborhood of many well behaved clients (the way it was when LimeWire had the only ultrapeers) you would have a lot better search results than you have now. That's why I think LimeWire should group their clients together.

If each servent vendor occupied a different area of gnutella, and if all the areas were well interconnected, it would be a lot more difficult for one vendor to poison the network for other clients. If all LimeWire / Bearshare / Gnucleus hosts were each grouped together, it would allow each vendor to implement new special features and experimenting without harming the network for others. Searches and downloads over the borders of the neighborhoods would be possible (although not every vendor's servent might be able to chat with another or show certain statistics) and if LimeWire turned out to return less search results then for example Morpheus, we would know, that it is LimeWire's fault.

The way it is now, when some servent seriously screws up, the gnutella network becomes bad for everyone.

just my 0.02€.

I agree, logically if you are a network administrator with 2000 users, you aren't going to be operating on a peer basis ... you're going to have at least one server somewhere ... probably several.

So basically, the more gnutella clients, the worse off the network is. There are a slew of clients listed at Gnutelliums , each with a different vendor, not to mention all the "home-grown" variations of the client.

If the vendors went to a pay-for client, things would improve. Any thoughts?

Ultrapeers are a good alternative to real servers. They appear and disappear dynamically to keep exactly as many ultrapeers as are currently needed by the network. Real central authorities make the network legally vulnerable.
The number of different clients is not the problem. The problem is that gnutella is an open protocol, anyone may write a client and there is no guarantee that all clients will be well behaved. So what LimeWire and the other clients need is some sort of defense against other clients and grouping is one approach to do that. Especially if a bad client is popular and makes up for a substantial portion of the network.
More commercial clients don't solve anything at all. Even if a client is commercial it does not mean that it's a good one.

http://mitglied.lycos.de/mdsgeist/LimeWire.zip

LimeWire version with the added ability to block gnutella connections to certain vendors.