What's up with upload preferencing?

No credit system, please. If somebody really wants to be a freeloader, let him.

I don't make the design decisions at LimeWire but I'm quite sure the LimeWire developers would agree with me there (for a change). Freeloaders are not a serious problem at the moment. People downloading lots of files are usually also sharing at least some of their downloaded files.
Partial filesharing doesn't work as well as I would like it to, but that's one of the points where I disagree with the LimeWire developers.

Thats is why other networks are more popular. Freeloaders are a serious problem, especially after many countries changed their laws - nobody wants to be cought... I believe that at least 50% are freeloaders (note at least), what do you think if these users would only share 2kb (of course high speed connections should share more) of bandwidth each, would this increase the network performance or not? I don´t need to be Einstein to see it does...

And I changed my minds within the years, at the beginning I thought let them download (who cares), but nowadays I say if somebody expect that a different person is risking to be sued for, the downloader should also risk something... if the risk is to high, go to the shop and buy the stuff!

And we shouldn´t forget freeloaders are people who share NOTHING, where is the problem to share at least legal files?

Morgwen

The percentage of freeloaders is something like 33% according to BearShare's statistics, so this is not a serious problem at the moment. You might want to try reducing that figure by adding some message to the GUJI telling users to share files but credit systems are not the way to go.

Not only are there more immediate solutions to increase the performance of Gnutella, - credit systems as a whole are not very effective in an open network. I don't want to go deeper into this topic because it is fairly late and I'm watching a movie, but I will say this much: The probability of meeting the same client twice within a short amount of time is very low in a large network. You would have to handle a huge database of information on who downloaded / uploaded how much. The number clients that would effectively ever be able to benefit from those credit systems would be very small.

The credit system would just produce an illusion of fairness, maybe have some users share more files because they think it allows faster downloads but you wouldn't even be able to measure the effects reliably.

33% not a serious problem? If you are a shop owner and only 67% of all people would pay would you call it a serious problem or would you increase the prices for ALL people? Up to 5% is no problem but more of course is...

Again my question where is the problem that these users are sharing legal files?

Its not necessary. A credit system based on the percentage of the upload speed you offer is enough. Of course this would only work with partial sharing, so the people would share at least the things which they download. See eMule!

Morgwen

Unlike edonkey, Gnutella has mainly small files. What you propose simply won't work. The probability of two servents downloading from each other simultaneously is minute (even for larger files, by the way). And without eMules awful, long queues this will not happen.

eMule queues are so long because most of the people share larger files - what do you think how long would the Gnutella queues be if the people would share mainly larger files?

Yes Gnutella is good for MP3s and nothing more... the Donkey net is very good for larger files and good enough for MP3s... a good all rounder. Perhaps Gnutella should be renamed to MP3 sharing net? This is one of the resasons why I only rarly use a Gnutella applications, it sucks for larger files and the freeloader problem.

I have to repeat where is the problem that freeloders share legal files? There is NO exuse...

Morgwen

sharers need to be encouraged--but someone can also share bandwidth by acting as an Ultrapeer (here I use half my alloted bandwidth of 1 gig per day just acting as an Ultrapeer).

[HAS the freeloader preferencing in LW's options been replaced by something more automatic in LW's code?]

They wouldn't be any longer than they are now because most gnutella clients limit the queue length to ~10.
The first-come-first-serve principle does not make file transfers more efficient so long queues are clearly not needed for a p2p network. (Rather the opposite is the case because long queues mean lots and lots of wasted bandwidth just for transmitting useless information like position in queue, next re-request time and so on.)

It may seem that way because most Gnutella users don't seem interested in large files - or rather Gnutella's selection of large files is small compared to edonkey's but gnutella can handle large files almost as well as bittorrent: The only problem with Gnutella is that most servents don't download the file chunks in random order yet, but that should change soon enough.

eDonkey is inefficient for almost any task. For downloading a single file it is very slow, the long queues and the source-exchange produce lots of unnecessary ovehead and it heavily depends on servers.
If you want to download a single unique file, from a single host, its long queues can be very frustrating unless this host has a very high uptime.

I think you can try to encourage users to share by showing them some "please share" messages. It is impossible, however, to force users to share in an open network, - or even create an efficient rating system that effectively disadvantages freeloaders (see eMule).

Ah that is the problem. You need MUCH luck to get a free slot and so its takes MUCH longer to download a larger file. With the Donkey my place is save... and with usually several hundreds of sources you won´t wait very long.

Yes because the people who want to download larger files use the donkey - why? Because of the problems I mentioned. gnutella will never be a good all rounder, to many cooks (developers). Did you remember the donkey started as one client, they made the protocol as they wanted... the users saw its good and used the donkey. But within Gnutella the developers have to discuss two years before they add needed features!

Bullshit!!! For larger files the Donkey is THE BEST WHAT EXIST, only with smaller files you "might" need longer to download than with Gnutella.

I never had this problem that I had to download a file from ONE HOST. The chance for anything like this is very low, you will have the chance to start the download from an other user who is downloding it at the moment (partial sharing). And it would me frustrate much more when I don´t get a free slot and I don´t know if I will get one even after several hours of uptime... the 10 queues limit is like a lottery, with the donkey I see "exatly" if its worth to wait or not. btw people who download and share larger files has usually a very high uptime or did you manage to download a 700Mb file in a few minutes?

Its better than ignoring the problem...

I think we should disadvantage freeloaders in any possible way, the use our bandwidth without giving back anything - they are disadvantaging the people who share their files - they need longer because some people don´t share their files. All sharers pay the price for some selfish freeloaders

As I said there is NO exuse to be a freeloader even a 56k modem user can share at least 1Kb of his bandwidth...

Morgwen

That is an illusion. Queueing has no influence on the upstream of a host. You need more luck to get a free slot in a queue but the average time you have to wait for a free download slot remains the same. Sometimes the you may get lucky and get a slot almost immediately, some other times you won't. Since the queueing model doesn't have any influence on the average upstream it doesn't matter for the whole of the network, if you have queueing or not.

Only an hour or two, - which is much too long, IMO.

I would rather call that "normative Kraft des Faktischen" to use a German term. The crowd goes where the crowd goes. It's difficult and not really my goal to convince all those happy eDonkey users to use Gnutella. I could give you more examples where large numbers of users kept using technically inferior networks, just because that's what everyone uses. Fasttrack is just one example, WinMX is another.

Spare me your collection of inappropriat proverbs, please. Your ridiculous claim is totally unbased.

Well, they had to face other problems than eDonkey because Gnutella development was not driven by a single team with a single philosophy, so development was slower, had a slightly different focus and encompassed a lot more experiments. In addition, eDonkey still has an expensive client-server architecture that is kept alive by the community while Gnutella is one of the few networks that can survive completely without servers. Even ultrapeers can operate with less than 10KB/s bandwidth in both directions.

I did not ask you to abandon your religious beliefs but I still tell you, eDonkey is inefficient and produces far too much overhead. With my 128/768 DSL connection I can upload ~10KB/s of real content before my downstream is affected on eDonkey because I don't have enough outgoing bandwidth left to create an insane amount of 5-7 connections per second and keep those hundreds of unused connections open just to stay in some queues. Gnutella doesn't have that kind of overhead.

Lucky you! The chance of downloading a file from ONE host is actually quite high when downloading mp3s because people tend to mess with the mp3-tags, re-encode files, share their own tapes from live concerts which are usually very rare on any network. And I have used eMule regularly, so I know how it is to have a file with one or two sources hanging in your download queue for weeks. Btw, we have had partial-filesharing on Gnutella for a while. The only major vendor who is still working on partial filesharing is BearShare.

On Gnutella you would very likely not even know its there. Most busy hosts don't return search results because it's a waste of bandwidth and all it would achieve would be having them bombarded with download requests which can be quite expensive (a simple TCP connection costs easily 100-200 bytes, sending the request is easily another 100-200 bytes. If you get 5 of those a second like on ed2k, you are bound to see some negative effects concerning the amount of content you effectively upload).

There is no easy way of disadvantaging freeloaders, unless you want to create a market for a specialized freeloader servent.

Just now I am downloading a movie (700MB), from Gnutella. Found 6 sources for my search and I have been downloading it from 2 sources at 20KB/s for two hours now. If you can find a large file on Gnutella at all, you will usually be able to download it, - unless you see the 'RAZA' vendor ID next to the search result.