Adam,
I don't doubt your a friendly and sympathic person. Unfortunatly, the decisions of your company do not implify telling the truth and technical needed decisions. Limewire definitely did not tell the full truth about spyware, and somebody from your company called me "spreading X-cases" to make me silent. It wasn't easy to discover the truth about your Spyware - and it's still bundled (not everything is opt-out) and will infect hundreds of newbie users.

About clustering again, that sounds like an excuse. The clustering is not originated at your "Ultrapeer" only, it also comes from your hostcaches too. For both I see no technical reason, perhaps your superpeer model is not very reliable and shold be improved. Let's go into detail:

> The reason for preferencing them is basically just that in hitting
> one UltraPeer with a query, you're in fact hitting up to as many
> as 80 (and in the future up to perhaps 500) nodes on the network

First, my horizon is much bigger than 80 or 500 peers now. Second, how will you improve your total accumulated horizon on that 6x size, without the other non-limewire clients involved? Your basic estimation must be wrong or will abuse other disadvantaged clients. It still sounds like a two class community for me.

> It's just a much better network model that improves Gnutella
> for everyone, and we're clustering them because otherwise you
> don't get nearly as much improvement.

You described that a superpeer model is better, yes I highly agree that dynamic network structures are what we dream from today. But you did not describe how clustering will improve that superpeer model. As I told before we do not need a 100% staturation of superpeer in a horizon, of course a small percentage is far enough. Don't forget normal peers users act as links in networks too and many can be grouped/shielded behind superpeers. We have a decentral Gnutella network, if you want a centralized system full depending on clustered superpeers, I which you good luck against RIAA. Yeah, the outdated centralized eDonkey superpeer concept works much faster currently (it has even more features as your "Ultrapeer" concept AFAIK), unfortunatly a network structure based on central servers or client monoculture can be easily shut down or attacked.

Don't forget that clustering is not fair and will create a two class Gnutella. Don't create selfish advantages for yourself. This reminds me so much to Vinnie's politics, do you think your client will win the run (especially with that market share you have today)? What happens if Xolox comes back again, with more astonishing technoloy as last time? Or disadvantaged devlopers will not tolerate that unfair behaviour in future and hold back their own improvements.... what you do is causing as split of Gnutella.

Greets, Moak