A New Peer-to-Peer Network

[iNaNimAtE]

Considering this is a forum all about programming, I thought it appropriate for this post.

For years now, I have seen peer-to-peer networks come and go. Each of them has their good qualities, but in each lies a tragic flaw (individually unique). There recently has also been some problems with the RIAA and MPAA coming in and busting up people for downloading (mainly the RIAA). While I could preach for hours about this issue, I am going to state clearly the topic of this post: I am interested in making a new peer-to-peer network (and, obviously, client).

Although it seems like just another failing attempt, I hope to include (enhanced) the great successes from each of the clients today, along with some new ideas, into a "mean-lean-downloading-machine."

This new network utilizes hash-verified downloading; no fake files. I am to the point of thinking of taking out the "search feature," and instead, all searching will take place on a website, and then the website will send the hash to the client to begin the download. Similar to the "ed2k://" link you can find on www.sharereactor.com that utilizes eMule. Hash-verified downloading also accomplishes another task; extremely fast downloading speeds. If there is one copy of a a file floating around, when someone requests it, each person who has it will add their bandwidth to the total (multiple thread downloading) to create blazing speeds.

At this point, it is beginning to remind me of eMule. However, the queuing problem that eMule has will not be present in this application. Instead of limiting uploading by concurrent connections, uploading will be limited by your upload speed. This will be determined by a speed test which will be executed upon installation.

Also, the entire network is, of course, decentralized. This seems to contradict a previous statement I made about "...all searching will take place on a website...," but there are ways around this. I have a basic idea on how the decentralization will occur, but it is still just an idea.

*The reason for this post: Does anyone want to help out?*
This is obviously not a one-person job, so post your comments.

[redhead]

Originally posted by iNaNimAtE

This new network utilizes hash-verified downloading; no fake files. I am to the point of thinking of taking out the "search feature," and instead, all searching will take place on a website, and then the website will send the hash to the client to begin the download.

Why not use md5sum instead of hash?

If there is one copy of a file floating around, when someone requests it, each person who has it will add their bandwidth to the total (multiple thread downloading) to create blazing speeds.

At this point, it is beginning to remind me of eMule. However, the queuing problem that eMule has will not be present in this application. Instead of limiting uploading by concurrent connections, uploading will be limited by your upload speed. This will be determined by a speed test which will be executed upon installation.

Hmm... why not make a test for "possible upload speed" uppon installation, then lateron the user can specify "I want to use 50% of possible upload speed" This way you wont have users complaining about extreem latancy, due to some one using their bandwith.

Also, the entire network is, of course, decentralized. This seems to contradict a previous statement I made about "...all searching will take place on a website...," but there are ways around this. I have a basic idea on how the decentralization will occur, but it is still just an idea.

Are you thinking of the client sending info to some server about md5sum of available files, with info on where these files are, which then on this server produces the download.

If you want, I could use some time on it, altho I would like to know what type of license a thing like this would be available under. Personaly I wouldn't put any time into things not worthy a GPL like license.

[joe_bruin]

Originally posted by redhead
Why not use md5sum instead of hash?

md5 is a hash.

[rdove]

You guys are getting me high talking about all this hash

[safecracker]

are you the same guy as from thebroken....if so you already know my thoughts.

[Belisarius]

I think he's confusing hash with crypt().

Originally posted by joe_bruin
md5 is a hash.

[iNaNimAtE]

Yes, I am the same person as thebroken. ( I sure hope no one else is using my handle). I think MD5 would be fine as a hash.
How the hashing works:
Method #1: Each person has a list of available hashes on their system. You use a main website to search and download files. This website sends hashes to the program. The program goes in search of people's hash lists to find the file.
Method #2: Same as #1, but there are supernodes who keep a master hash list of the shared files for 100 users or so.

Redhead: The bandwidth idea was good. I like it. This program would probably be under a GPL license. It would obviously be freeware, and no advertising.