Here's my design. I would love to know if you decide to develop it, but the following liscence applies:

1) If you develop arround this network model, you cannot claim origination of the ideas
2) If you develop arround this network design, you cannot make it propriatary
3) If you develop arround this network, I would love to know about it :-)

Anyways, here is my ideas...

Level 1 Ultrapeer
- Contains the MD5 (possibly bitprint) Hashes of All Connected Level 2 Ultrapeers and their nodes
- Nominated by Level 2 Ultrapeers
- Maintains full routing information for all Level 1 Ultrapeers and Communicates with them Often
- Extremely rare files for which nodes are aware of the MD5 Hash will have their queries distributed among the Level 1 Ultrapeers.
- Level 1 Ultrapeers will collectively loadbalance the entire network based on preset methods
- Level 1 Ultrapeers can at any time be "voted off" by election of other Level 1 Ultrapeers or multiple Level 2 and 3 Ultrapeers... This ensures that a Level 1 Ultrapeer that is not behaving will be quickly pulled from service
- Multiple failsafes MUST be inacted for Level 1 and 2 Ultrapeers

Level 2 Ultrapeers
- Contains the MD5 and Filename Hash of All Connected Level 3 Ulrapeers and their nodes
- Nominates Level 1 Ultrapeers based on Variety of Variables (Uptime, Speed, Ping, etc)
- Has multiple Level 1 Ultrapeers through which it sends queries searching for additional hosts for a file and recieves a routing ID from each
- Connects to at least 5 other Level 2 Ultrapeers to send Queues for Files
- Is Nomindated by Level 3 Ultrapeers AND/OR other Level 2 Ultrapeers

Level 3 Ultrapeers
- Containts the MD5, Filename Hash, and Full Metadata of all Connected Nodes
- Responds to all Queries directed at its nodes
- Nominates Level 2 Ultrapeers based on Variety of Variables (Uptime, Speed, Ping, etc)
- Has 3 Level 2 Ultrapeer Parents each giving the Level 3 Ultrapeer a Routing ID
- MAY Be Firewalled

Nodes
- Has "Parent" Level 3 Ultrapeer through which ALL Network Functions are Performed (except Upload Queue)
- Has 1 Additional "Alternative" Level 3 Ultrapeer through which "Search More" Feature is Supported (Search through this Alternative
- Maintains Upload Queue
- Maintains full Alternate Locations of files Node is currently downloading or reciently downloaded (NO IPs!)
- If node is not useing both upload and download bandwidth, the Node can be given "Jobs" by the Level 3 Ultrapeer (if node settings permit) such as Proxy for 2 Firewalled Hosts OR given files that are highly requested to help distribute them


Defintions-
Routing ID:
This is a mathamatically equated number that is the primary way a message is tracked accross the network. Based completely on the parent nodes through which the network service is sent through. Each node and each Ultrapeer should only have 1 main Routing ID to give to those requesting files, but may have multiple (as needed) for network services. Level 1 Ultrapeers can have 1 (and 1 only) Routing ID, and the entire Level 1 Routing Table should be published at regular intervals throughout the day to all Level 2 Ultrapeers, and twice daily to all Level 3 Ultrapeers. You cannot figure out the IP address by looking at the routing ID, but you can easily find the pathway to that Routing ID if it is needed. The routing id should be transmitted similar to this: Level1RoutingID:Level2RoutingID:Level3RoutingID:No deRoutingID

Upload Queue:
The upload queue is not based solely by the order in which the request is sent into the client. In fact, only 40% of the placement in the queue is based on the order. Other factors such as connection speed, percentage of upload speed sharing, upload/download ratios, uptime of node, popularity/rarity of file being requested, and the importance of the file (Low = 1, Normal = 5, High = 10) transfer to the client. The popularity and rarity of the file (or portions of the files) will make the biggest difference because a rare and popular file is the most important file to transfer. Also, very small files (under 100k) should be transfered almost immedeatly, but a cap on the number of small files that can be shared immedeatly (per client) must be set (to avoid abuse).

Filename Hash:
This is a hash of the search keywords for an individual file. For a music file this would be the hash of "Audio Artistname Songtitle", for a Movie it would be "Video TitleOfVideo". Only when a search keyword is positive to a queue will the Level 2 Ultrapeer request the Metadata for that file from the Level 3 Ultrapeer (as well as the ID of the node(s) that have the file).


IMPORTANT ADDITIONAL FEATURES NEEDED:
- Partial File Sharing
- Full Metadata Support


NOTE:
If a client cannot support all the features listed above, it should not be allowed onto the network in mass quanities. Educational and research is fine, but when 50,000 people start using software that won't let them use their computers to the maximum, we have a problem.

NOTE 2:
You are not requred to share files on this network, but if you don't you won't be on the top of the line when it comes to download queue. Quality of Service for those who do not share upload bandwidth will also be deminished (don't expect to send out a ton of queues).

NOTE 3:
The network needs to be able to build on all new technologies, including wireless mesh networks and 3G wireless phones and PDAs. Thus there is no need for a node to take ANY active part in the network because odds are the bandwidth in those cases just isn't present.

NOTE 4:
Don't be strict to the HTTP specs, nor be strict to IPv4. IPv6 provides a ton of benefits to P2P networking, so don't lock the technology out before it is even deployed.

__________________
- Nick Catalano

Your use of hashing doesn't seem very useful. Hashing the name of the file doesn't seem like it would be all the useful.

The other developers won't take this serious anyway,
or why do you think no one replied?

I think speaking about coding is easier than doing...
and obviously someone here has no idea what he's doing
(WHAT IS HASHING FILENAME GOOD FOR?)

__________________
Never touch a running system.
www.guox.de

Of course the entire file should be hashed,
what else is hashing for file validation good for?

__________________
Never touch a running system.
www.guox.de