Hi!
"Cooperation can flourish if the public-spirited majority can punish freeloaders, say Swiss economists. People will pay to punish - suggesting that their notions of fairness outweigh selfish considerations. The work may help explain why people cooperate in society." tells a new Nature article:

http://www.nature.com/nsu/020107/020107-6.html
"Prosperity through punishment"

PS: Funny. Personally I do not like punishment, but giving gnutella members a feeling of "we do not tolerate freeloading" and "punish" them with friendly anti-freeloading countermeassures" (force swarming) and more control (chat/showing generosity) could also improve cooperative behaviour = better sharing. It's an emotional aspect, the article above could improve our understaning of human cooperative behaviour.

PPS: Also a Slashdot and infoAnarchy article now, see:
http://slashdot.org/science/02/01/10/0324223.shtml
http://www.infoanarchy.org/?op=displ.../11/02517/7790

How can you tell if someone is a freeloader when they start downloading? You can't right now.

Personally I would simply like a few extra http headers, and without those a download will not occur.

eg:

Hosts: 8/10
Leafs: 0
Shared-Files: 1440302342/435

Shared-Files is usually the only thing one might wnat to check.. the rest would just be nice for ultrapeer **** :P This would then tell a host whether or not somebody is a freeloader. If all clients keep this _right_, then one can simply reply when one is considered a freeloader:

HTTP <error> Freeloaders aren't allowed to download

or whatever... Thoughts?

That could be faked

Add to the protocol data about:

* number of files shared
* number of bytes shared
* ratio of downloaded/uploaded bytes (lifetime)
* highest possible upload speed
* current minimum upload speed guaranteed

Another method used by some p2p programs:
Ask the user if you may automaticly search for media files, and then share folders with a lot of mp3, avi, etc. files. Most users will just hit yes.

OK, stop being control-freak nazis.

There are many things which already make gnutella a problem Lack of files is not one of them.

There are enough files.

The problem is it is too hard to find the ones you want ..

Fix this by:

1. Support reporting of bitrate, song length etc
2. Allow limiting searches based on 1 (to reduce unnecessary network traffic)

3. Prevent reporting of shared files whenever there are no upload slots
(this is currently a BIG problem I think with limeware - because in the current version if you set upload slots to 1 it is in fact read by limeware as 0!!)

4. hashes. Not swarming, which sounds very difficult to implement. If you want swarming, go to freenet ;-) I would not use a swarming client because I am on a modem connection and i only want to spend bandwidth on files I want. I share heaps, I don't want some nazi making me share stuff I couldn't care less about.

Hashes will make it so much easier to get a download to finish, because you will be able to continue a download easily from another source. Clients will be able to automatically search for alternate sources and pick up a faster/more reliable one.

5. Hashes enable what I would call swarming: ie client with enough bandwidth downloading from multiple sources different parts of the file. Implement this instead.

6. Clients should allow user to prevent uploads to freeloaders .. either based on number of files shared, or based on the existing 'did upload' or whatever. But it should definitely be optional and at the user's discretion. But don't go overboard and implement any new stuff in the protocol to achieve this. Make use of what you've got.

7. Remind users when there are still people dling if they go to shut down their client.

8. Make clients which consist of a daemon and a front-end, with the daemon always running (and sharing) - but ask the users before setting it up like that, because that can really annoy some poeple (like sysadmins).

9. Genorosity indicators suck. I don't want to be labelled as a freeloader if I'm not one. Like other people here, I share heaps, currently > 13 G, but I am on a modem so not many uploads very often. But 11G of my stuff is 'original' (ie I ripped it myself, not downloaded from gnutella). As any genuine gnutella user will tell youi, it is the hard-to-find stuff that is most valuable, not the stuff that every man and his dog has downloaded and deleted a million times.

He he - the people who will get the best generosity marks if you implement it will be those suckas with the gnutella virus! How funny - they would be treated like gods in your 'community' because they have little virus protection and less brains.

** End of list

Really the worst thing about gnutella is finding files, downloading 90% of them, then sucka at other end cuts you off and you never find that source again. Fix those issues FIRST. Then you can make it into a game, but I won't play, thanks. If I want to play a game I'll visit my broadband friends and play counterstrike. It's more satisfying.

Oh, yeah, and as a modem user I am tired of seeing this hype that ultrapeers will save me so much bandwidth.

My download bandwidth is never a problem - I will get 4-5 k a sec from a good source, most sources (even 'fast' ones) don't run more than 2-3 k/s. Limewire does use less bandwidth for connections, but I don't think it it helps the downloads. Anyway, the point is, with 5k/s available, and two sources running at 2k/s, that still leaves 1k/s for connections, which is plenty, given that the download is gonna take like 1/2hr to 1hr. I get plenty of search results over that time, and can happily upload and download at the same time. Really, stop it with the 'ultrapeers will give modem users so much more bandwidth. It's not like they're gonna come around here and plug me into cable, you know?

I think up's are great for increasing search results, but that's the primary benefit. Not bandwidth for modem users.

oh yeah, number 3. needs to also not say there are shared files if it is patently obvious that the client is unable to successfully upload.

This will require some extra protocol work, I think, to really get it going, or something (I don't know enough!) but basically, there should be a query 'please test my upload ability'

Send that query to another client and the client requests some file from you (doesn't have to be a real file, could be some special filename like 'gnutellatestYYYYMMDDHHMMSS' that you told it you had, but actually you produce the test data automatically)

If you try a given number of clients (say 10) which say they support this protocol, and you don't even upload 10 k to them, then you don't go telling the whole world about your 5Tb of unique mp3s (which no one can successfully get because your sysadmin long ago blocked all incoming queries).