Quote:
Originally posted by SA_Dave
I may not have posted what I meant in point 7! I shouldn't have used the word "deny".  What I meant was, that SN's should be able to decide what files are maintained in its database. For example, I could 'host' anime episodes, pix & mp3's for example. If a standard peer wanted software, I would redirect them to another SN with full access to that content. However, if a user only wanted anime-themed content, I'd only search my peers & any other SNs that have similiar content. This wouldn't mean that any peer is banned as such & in this way it might improve network performance if you share a diverse collection of files (you connect to even more SNs) as well as limiting search traffic of those who want specific content. It might also provide a filtering mechanism before the search results are returned. I don't think the direct connect style of private, invitation-only, or strict-requirement groups will work. These "groups" would remain public, but it should be done in a transparent way. If my db idea were implemented, a SN could also link to specific files hosted by other SNs without content preferences, thus limiting search results to specific files only. This would be extremely difficult to do, I know, but it would be a nice feature. 
|
Hi
SA_Dave and welcome onboard - enjoy your stay at Napsterites!
An excellent post and a pregnant idea which resonates well with my own thoughts on developing the connectivity and scalability of p2p networks.
The ideas of supernode architecture and multisourcing have taken p2p to a certain point but this point is far from satisfactory both contentwise and socially. Instead of providing us a full connectivity and full search visibility among millions of concurrent users
all today's decentralized networks open up just a very limited part of peer space for our social presence and searches. My gut feeling is that I see and am seen by perhaps 10.000 other peers on WPN and my last experiences of FastTrack were rather similar. Instead of having the present 1 % visibility in a million-strong p2p community we would really like to have 100 % visibility - to have working hotlists and communication channels to all our friends (like the instant messangers have) and to reach with our searches - even if slowly - the entire shared content of the network. Considering how fast the p2p communities have been growing the design target for the next generation 'decentralized Napster' should really be around 10 million connected peers. That's perhaps 4 magnitudes from where we are now so there is plenty of room for architectural innovation.
Your example demonstrates one way to let the
content of the network to shape its connection topology further from the initial 'random' distribution of the supernodes and peers subject to them. The generic goal of getting specific content and people who like it to gravitate closer to each other in the network topology can also be achieved in a more autonomic, self-organizing fashion. The supernodes can use the information of successes and failures of long-distance (supernode-to-supernode) searches and related downloads as a guide in reorganizing their respective peer populations. It makes a lot of sense for a reggae lover to find his/her way gradually closer to other reggae lovers - under the same supernode or into the same group of effectively clustered supernodes capable of linking the interest/genre group together. The beauty of the self-organizing scheme is in that it would not be bound to any predefined genres or themes or fixed supernode hosts. It would rather clusterize any peers who like what they find from each other's libraries, whatever name they want to give to their common interest. And as a peer, all you would have to do to get closer to better and more interesting peer sources would be just to keep searching, sharing and downloading the stuff that you personally like most...
Very few of us are limited to just a single genre interest and some of us are very eclectic. A content-shaped topology would probably work best with a more multidimensional supernode architecture than the present exclusive one. Rather than being connected to a single supernode in a single content cluster you would probably want be connected to a number of supernodes, each serving and representing a particular genre/community interest. The subjective experience of this would perhaps be like having a house in cyberspace with a number of doors opening to very different streets with different local communities and content pools on each of them.
Quote:
Originally posted by SA_Dave
I apologise for the smilies. I just got carried away I suppose and didn't realise this was a problem?!?
|
Do not apologize! Your first post was brilliant both content & smiliewise!
- tg
