Slashdot Mirror
The Gnutella Paradox
bemis writes "Red Herring is running an article about Gnutella and how its success may ultimately cripple it ... also covers the background for the uninitiated (like much of the 'management-types' that read RedHerring) of Gnutella, and Nullsoft itself." The article covers a lot of ground and is worth a read.
It seems like all of these "truely" distributed will all have the problem of scaling poorly due to the large bandwidth overhead caused by things such as push requests and pings. Why not create a system that, although not truely distributed, doesn't really have to be as centralized as Napster?
Why not model it after DNS? There could be, oh, several dozen "root" servers that keep records of all the IPs that list where you could get files that have a certain file extension, (.zip, for example), much like the root servers for DNS, which list all the servers that serve the various TLDs. Then, when the requests reach the servers on the next level down, you recieve a list of IPs of servers that have a list of clients that have the file you are looking for. Like the DNS system, all the locations of the files you downloaded would be stored in "your" server, which would be on the bottom level of the hierarchy. The clients could be programmed to log into a different lowest-heirarchy server each time they log on, thereby providing some kind of load-balancing, so that no one lowest-rung server has significantly more clients connected than the other.
Although this would throw anonynimity out the window (with all the IPs cached somewhere), I think it would scale a lot better than Gnutella, since it would eliminate a lot of the traffic on the network, or at least channel it to the machines running the server application.
IANAP (I am not a programmer), but I know that a lot of Slashdotters are, so I am counting on everyone to let me know if I am either talking out of my ass, or if I have a genuinely good idea here.
Friends don't let friends use multiple inheritance.
this post is for the information have-not's in the press... gnutella is not going to be 'the new home of napster fans' ... openNap will be.
I just fired up gnapster, and chose to 'browse OpenNap' servers, I have damned near 100 to choose from. All of these servers are indepenent, not related in any way to "napster" inc.
Sure, right now the song selection on these servers is pretty crappy, but following the mass exodus I'm sure some snappy hacker out there will write some sort of 'bridging' software to bring all these other servers together,
you poor media types just don't have a clue, do you?
Isn't it obvious that the best benefit of using something like Gnutella is the possibility of setting up "private" Gnutella nets?
Sharing with the whole world is always a problem. If you've got something everyone wants, they will always beat a path to your doorstep... with a number of friends in a "private" network, the grass might not all die as they cross your yard.
-Nev
-Nev
> but if Gnutella, which has some of the best open-source programmers
> on the Net behind it, can't survive the technical or legal challenges
> of critical mass, how will the other programs be any better prepared?
Mr. Fenning created Napster with the help of a few people.
Justin Frankel created Gnutella with the help of a few people in a few days.
The other programs don't have to be better prepared. They simply have to be better.
We've seen the best application that the Net has yet to provide.(*) We'll never forget what can be. What should be. What must be.
I'm itching all over, and some day, I'm going to scratch.
(*)- Admit it, when you first used Napster you said to yourself: "Holy shit, this thing is fast and does EXACTLY what I want it to, unlike 99% of every other piece of software and place on the net. This is EXACTLY how the world should be. This is spectacular."
I think this account of GNUtella is obvious, but long overdue. Though I've been saying that GNUtella was flawed since day 1, it suprises me how much of slashdot used the "but there's always GNUtella argument". What's more, though somewhat less suprising, has been the tendency of the popular media lately to take this same attitude, acting as if massive piracy is necessarily unstoppable because of the idealized GNUtella.
One thing that I'd like to add to this argument against GNUtella is that, even if it or any of it's cousins could scale, I believe it'll ultimately fall due to the combination of two primary factors, the so-called "tragedy of the commons" and the threat of law. In other words, it's already been established that damn few people are willing to really share. Today, the only primary reasons not to share seem to be bandwidth and CPU concerns.
In the future one thing that will play a dramatic element in the mix, if RIAA must, is fear of lawsuits by the fileservers. Since the beginning of the spread of piracy online, there has been very little fear of enforcement. Put simply, neither the software industry nor the music industry has done much more than attempt to threaten the owner or the uplink of the server. However, if the industry starts prosecuting (or otherwise hurting) the top fileserves, that will mean that anyone who runs a server is putting themselves at risk.
Put it this way, as redherring alluded to, there is very little reward for sharing files. Yet, there are costs (i.e., bandwidth, cpu/hd utilization, and time). So here we have a new emerging cost, the threat of getting busted. If they go after the top 1% of fileservers without fail, I think it is reasonable to assume that most everyone will make sure that they're not in the top 1%. Thus the entire organization will collapse from the top down.
It need not even necessarily be criminal prosecution. If the industry can effectively get a user blackballed from all broadband ISPs, how many users are going to risk it? Think about how many of these servers are either being run off of DSL, cable modem, or universities. I can tell you that most of these authorities can, will, and have, rapidly shutdown offending users services. Those users that share the most, are also most likely to not cope with losing high speed service. It's a significant threat.
The cost benefit ratio for the file servers is very poor as is, so low that I doubt it'll withstand serious additional pressure by industry. I know many of you will clamor, "but one will always pop backup"...but this is flawed logic. Why would any rational individual pop up and make himself a target?
Though it may be true that COLLECTIVELY if no one runs scared, then those industry tactics would be ineffective. But this is thinking too much like a group. It's like saying that automobile traffic would be infinitely better if everyone aheared to a few basic rules (i.e., no cutting other drivers off, left lane for passing only, etc). That is very true, but that doesn't mean that it is better for the individual in the driver seat at that moment. At that moment, the individual is thinking like an individual, not as a group, so, traffic happens. Likewise, "tragedy of the commons" will happen.
I can easily see RIAA coming after all of the top fileservers (no matter what the protocol). So long as the uploader/fileserver is known, they can be targeted.
Mojo Nation is a p2p file sharing protocol that has a built-in digital cash system. It prevents the "Tragedy of the Commons" problem by effectively creating a barter system for bandwidth, disk space, and CPU. In order to search, upload, download, or otherwise consume any resources from the remote host you must compensate them with the internal currency, known as Mojo. The Mojo represents the resources you are consuming from the counterparty. This way nobody can consume more resources than they are contributing to the system. Each person who joins helps to make it stronger. Note that contributing resources doesn't mean uploading files. You must pay Mojo to upload since you are consuming other servers disk, bandwidth, and CPU by uploading blocks to them and asking the servers to hold them.
The best way to get Mojo, so you can get the files you are interested in, is to provide your own resources (bandwidth, disk, CPU) to the network by using the Mojo Nation Broker (our name for the client software) to run a Block Server, Content Tracker, or Relay server.
A Block Server holds the actual data. In Mojo Nation, instead of holding an entire file on a single server, every file is broken up into many redundant blocks which are spread over many block servers throughout the network. You only need half of the available blocks to reassemble the original file. Of course, the Broker does all of the hunting for and reassembling of blocks transparently. In this way Mojo Nation is like a big distributed RAID drive which makes it resistant to servers disappearing. It also spreads the load out over many hosts, so when you download you are not impacting any single host or network connection severely (expect perhaps your own). It also means that hosts with slow net connections can hold data since each block is pretty small. Your Broker can download some blocks from slower servers in parallel with more blocks from faster servers. The Broker keeps track of performance statistics for each host so it can make intelligent choices about where to purchase blocks from.
Content Trackers are like the search engines in Mojo Nation. Instead of routing all searches through the entire network (which is what is bringing Gnutella down). Mojo Nation has centralized content trackers, but anyone can run one. The content trackers store rich XML metadata describing the files so you can easily search on different fields. The metadata also holds the instructions for your Broker to find and reassemble the blocks that comprise the file. So if you run a Block Server but not a content tracker you cannot know what data you are holding.
Relay Servers are for people behind firewalls. Mojo Nation is an asynchronous protocol. Relay Servers are used so you can send a request to someone behind a firewall. The Relay server holds messages for the clients to pickup, in exchange for some Mojo of course. Relay Servers will also be used for Digital Mix untraceability, much like the old Cypherpunk remailers.
In any event, it is extremely cool and is definitely worth checking out
Burris
First, many people such as myself have downloaded music from Napster. But the type of music downloaded is important. I have downloaded songs from my past. These songs are important to me because they represent my youth. But now I have a gig or so of my youth, and that is all I need of that type of music.
New music I do not dl from Napster; I stream that from mp3.com. THis music is from unknown artists. When I compare it to new mainstream music, music which must be paid for, I find it to be just as good.
That is because new mainstream music does have its hooks in me like the old music. And BTW, this is the real intellectual property insofar as music is concerned: connection to the listeners' past.
But all these former dynamics will be swept aside by the new Net distribution model. I encourage all of you to try streaming some of the new music from www.mp3.com, especially the electronic music; often this is made by one person on a computer. It's fantastic! Also check out all the foreign ethnic music offerings. It's all free and legal--and very good.
eat shiat and bark at the moon