Slashdot Mirror
Death of the P2P net Predicted! Film at 11!
87C751 writes "Cnet has a preachy, whiny piece bemoaning the peer to peer "phenomenon" and its lack of commercialization potential. The humor comes when they claim that bandwidth limitations will ultimately doom P2P (as though bits that traverse through a server somehow take less bandwidth than bits sent from one box directly to another). " Alright, I'm a little softer then the submittor, although I agree with some points. The area that I do question is how much is actually shared - most of the people I see out there are taking, not contributing to the Gnutella and the like.
If peer-to-peer does go away, it will be because service providers (telcos running DSL and cable companies with their cable modems) restrict upload speeds to intolerable levels.
At my home, upload speeds are limited to 112kbps (about 14kB/s). It isn't terrible, though uploading a single MP3 can bog down your downstream rate too (since the two-way handshaking that goes on takes up a small but significant amount of the TX/RX time).
--
Yet another set of commercial ventures predicts the death of something the use of which they can't figure out how they can charge money for.
Oh, and I really like how CNet's ``printer friendly'' version of their pages removes the graphics that are associated with the article but leaves the banner ads. Pathetic.
--
CUR ALLOC 20195.....5804M
What's needed is some P2P at the high end, to allow for seamless mirroring and reporting, with the lower end being more heirarchy based.
That way (for instance) you ask your local server for a file and it uses a P2P method to retrieve that file across the network of 'big servers', then sends it to you. That way your dial-up connection isn't slowing down the large scale network.
_____
My Journal
If you're transferring files from a friend and you're a bit bursty adn so are they, you end up downloading a lot more slowly than you could in theory. (imagine that you spend 30 seconds each minute running at half speed and so do they, you only spend (on average) 15 seonds each minute downloading at maximum speed).
If you've got a high-bandwidth cache in between you can get a 50% higher throughput. Also, if the file has been cached beforehand (because someone else downloaded it) you can get it at your maximum speed.
_____
My Journal
Ah, you are right in that they are different data streams. However, if I open, say, Napster on my cable modem, within minutes there are probably 10-20 users downloaden from my box.
/. or do other usefull things :-) my HTTP-requests (upstream) do not get through, or at least very slow (like, it takes 10 seconds to reach even my local ISPs page). Same goes for request/ACK/whatever-packets that are sent, also while downloading! This will slow down your download quite a bit.
Then, when I try to browse the net, read
So while you are theoretically correct, it doesn't entirely work that way.
Every expression is true, for a given value of 'true'
These days the attitude seems to be that unless something is a business model then it isn't viable. These P2P services will be around forever now that they exist. People will realize that each individual can pay for a portion of the network-bandwidth-server infrastructure required to maintain the P2P service without requiring a corporation to control it. Of course the people with the least amount of resources (bandwidth for instance) will take more than they give, but I think the system (I guess I'm talking in abstract terms here about some sort of ideal or at least workable P2P service like Napster or Gnutella) will tend to balance itself out. To each as they are able. As broadband connections increase more people will leave their client/server running all the time. I for one haven't seen a problem on the networks, have you? Just because a corp can't create a business model to support some astronomical farce of an IPO doesn't mean something can't be successful.
If it was so obvious, why didn't someone do it three years ago? Seemingly minor or incremental improvements in the usability or packaging of existing technology can be a breakthrough if the result is that hundreds of thousands of people suddenly become able to do something which they want to be able to do, but couldn't previously.
I suspect you have a narrow technical definition of what you think constitutes a revolutionary breakthrough. The fact that the recording industry is shaking in its boots right now is proof enough of the revolutionary nature of P2P file exchange. And it's this specific application and incarnation of P2P "technology" that the CNET article is about. Not that I agree with the article itself - I'm simply reacting to your unjustifiably dismissive comment.
Bits is bits is bits.
Uh-oh, Nicholas Negroponte is posting on Slashdot now!!!
How much is being shared? A whole hell of a lot. Where do you think all the people who "take, take, take" are taking from?
Here's some stats for my MP3 sharing over HTTP only. This doesn't count what I share on OpenNap.
dwarf:/var/log/apache$ head -1 access.log127.0.0.1 - - [22/Oct/2000:06:26:35 -0400] "GET
dwarf:/var/log/apache$ tail -1 access.log
XX.XX.XX.XX - - [27/Oct/2000:10:27:36 -0400] "GET
dwarf:/var/log/apache$ grep -v '^127.0.0.1' access.log | grep '\.mp3' | wc -l
503
(I changed the IP address to XX's to protect the identify of the person who made that last request.)
503 mp3 file transfers (some of which are partials and resumes, of course) in 124 hours -- or about 4 per hour. And that's a very small number compared to the activity I get on OpenNap (which I don't log, currently, but trust me -- it's much more than 4 per hour).
Those of us who share may be in the minority, but we definitely exist.
Ratios... points... offering services to earn virtual currency which you can spend to download information... sounds like a pretty good idea, eh?
That's what the Mojo Nation folks thought, too.
Of course, one of the biggest mistakes they made with gnutella is not putting in some way to block dialup users. There should be some kind of bandwidth checking in the connect code. Maybe have it send 50K of random data from the node being connected to which then times how long it takes to get back the CRC of that data, more than 3 seconds and no connect. That would cut out all the slow users and keep ping times between nodes to a minimum.
Hmm... maybe I should go hack the gnut code and put up a high bandwidth only net.
when I suck stuff down, I usually block/disable uploads, because I need that bandwidth.
I should probably turn it on at night when I'm not using the machine to give back, but I haven't bothered.. there's no penalty if I don't do it, so why should I? I know I'm not alone in that line of thinking, though it may be wrong.
BilldaCat
Upload and download are different data streams at data-link (modem) level, not at the transport level: TCP is quite chatty and all those ACKs have to go back somehow. If your connection is not to the same computer the upload is going to, ACKs cannot piggyback your packets and have to be sent as stand-alone packets. These packets will have precisely 1 bit of useful information, but still use "minimum packet size" bytes. Quite wasteful!
The articles on freenet's architecture are a bit vague, but it looks like they do something like a space/time tradeoff to achieve higher performance -- more space is dedicated to data and metadata (by duplicating it across servers) so that the path to any datum is shorter.
Finding stuff is a different matter, and I suspect part of the solution here is to learn to accept imperfection by design.
Yes, I see the Freenet design implements a finite TTL on a request. Combine with caching data it means the network adapts to more popular data, serving it across fewer hops. So the effective network radius you have to search is limited for popular data, and TTL limitations puts a hard limit on less popular data searches. This caps the growth of bandwidth usage per user, which is a good thing, but it also means that you can't deterministically find something that actually exists.
It's anyone's guess to how performance and reliability would be affected by scaling to, say, a good fraction of the current size of the web; I expect there will be some interesting chaotic phenomena that will be uncovered with respect to the precise way parameters such as TTL and cache size are tuned.
Anyway, my hats off the freenet people -- they're in for an interesting ride.
Post may contain irony: discontinue use if experiencing mood swings, nausea or elevated blood pressure.
A good lawyer can take down any published set of servers.
The freenet architecture is interesting because it is even more decentralized and the servers are networked to share metadata. This means that in addition to taking down a number of well known servers, the lawyers will end up in a netherworld where metadata is passed in a highly connected and nondeterministic way.
Post may contain irony: discontinue use if experiencing mood swings, nausea or elevated blood pressure.
as though bits that traverse through a server somehow take less bandwidth than bits sent from one box directly to another
The aggregate bandwidth needed for file transfers won't change; it's the bandwidth required for meta-information -- catalogs, searches and search responses, that goes up.
Has anybody done any theoretical research here? I'd guess that in a P2P network the bandwidth required to carry meta-information would go up O(N^2) -- that is you want to have a network of information distributing nodes that is some fraction of a complete graph. The Napster architecture, while introducing a single point of failure (at least from a legal standpoint), seems closer to optimal from a purely technical standpoint -- it centralizes meta information allowing O(N) growth of query bandwidth in nodes, and decentralizes data transfer.
Post may contain irony: discontinue use if experiencing mood swings, nausea or elevated blood pressure.
My current favorite MP3 sharing program is Audiogalaxy. It has a security and anonymity oriented design, but on the discussion board people are boasting about how many files they are sharing and how many gigs they've shared so far. Contrary to the GNUtella experience, most of these folks seen to be taking advantage of anonymity to share more, rather than less. Of course that could change.
"How perfectly Goddamn delightful it all is, to be sure" Charles Crumb
i remember talking with my father in 1992 about this whole "internet" thing. he thought that no one would be able to make money on it, and that there is no compelling reason for it to be used.
and then came the web and all hell broke loose in 1994.
now we've got a different protocol, one that keeps true to the original intenet of the internet, and allows "Peer to Peer" sharing.
geez, the internet has always been peer to peer sharing, this is just allowing us to go back to this philosophy, and allow everyone to truly contribute back, instead of only those with large amounts of cash needed to generate hits.
so, all of a sudden, we will be back to the model that allows anyone to communicate with anyone else.
We're taking the power back with P2P. Using the internet what its meant to do - communicate, not make a buck...
tagline
... hi bingo
Ok, here was my idea for a whizbang P2P architecture.
What do we need? Something relatively anonymous. Something relatively stealthy. Something relatively standard, familiar, easy to use.
So I says to myself FTP. Just hack on some extra functionality that allows pseudo-links to *other* FTP servers. Clients would traverse the filesystem and not know that they were actually getting listings from N servers away (much like Gnutella, but an FTP interface). Hey, why not slap on a new command, say, REVERT, which, when passed with a secret key, makes the FTP server revert into "dumb" normal FTP mode. Great.
So then I look up the FTP spec. And what do I realize? FTP *already has defined a seperation of control and data flow*. FTP *already theoretically supports proxying*. FTP *is* Gnutella effectively. Somebody please read the FTP spec, and implement a server which will transparently do proxying like this (the nested remote filesystem stuff would be nice too - not sure if that is specified by the RFC).
This has the nice added feature that any law that attempts to strike this down, will have to strike down the FTP protocol...it will then be laughed right out of court.
It's 10 PM. Do you know if you're un-American?
The P2P system in my sig encourages users to share by use of a distributed karma system (a little bit like a certain web news site ;)
The release this Sunday will have file sharing enabled.
0.02,
Mike.
Tales from behind the Lagom Curtain
I remember when I was into Hotline (www.hotlinesw.com) there use to be lots of free servers, then folks figured out that you could make people go to a web page and click through on a bunch of ads to get the password for the account that would allow downloading. Most of the big sites went to that model quickly.
---
DO NOT DISTURB THE SE
Seriously, come on guys. A couple months ago all the pundits were telling us how great P2P was going to be; now (at least some of them) are telling us why it won't work. RIGHT NOW ALL THEY ARE TALKING ABOUT IS VAPORWARE!
Why can't the pundits wait until they have something to talk about before they start talking.
Oh wait, it's because they are pundits.
Tarsnap: Online backups for the truly paranoid
These things do favor a server-centric internet over peer-to-peer connections. The common user is supposed to be a content consumer more than a content producer ( well, honestly this is quite true for 99% of users - including me ).
Ciao
----
FB
The Napster architecture, while introducing a single point of failure (at least from a legal standpoint)
Napster's centralized server is not a centralized point of failure thanks to OpenNap.
Will I retire or break 10K?
Oh yeah...then we'll just have some jackass uploading Britney Spears mp3's renamed just to get download points...*sigh*
--
Wooden armaments to battle your imaginary foes!
The area that I do question is how much is actually shared - most of the people I see out there are taking, not contributing to the Gnutella and the like.
For my own part, I think there is a certain feeling of "it's OK if I take, but if I share, I'll be caught" It's the diference between finding a $10 bill on the sidewalk and intentionaly shortchanging someone. The guilt level is oh so much less.
Dirty Pirate Hooker
It all depends on the application.
.sigs??
Peer-to-peer works well in some instances, and star networks (server) works well in others. Just because more people see star networks working more often than peer-to-peer doesn't mean they are going to die.
Two examples:
1. Directly downloading a file from a friend. Peer-to-peer is by far the fastest way. The server would just be a middle man slowing you down.
2. A FPS game would be incredibly laggy on a peer-to-peer because of the size of the overhead. The star-network is the better choice here.
If you think about it, the internet is kinda both a star and a p2p network. There is no 'one central server', but a Peer-to-peer network of servers. So the P2P type of connection is going to die? I don't think so...
-- Don't you hate it when people comment on other people's
Good quote, too many chars. Seriously, the slashdot 120 char limit sucks!
I think that alot of people don't share on Gnutella, because they honestly don't know how. Once, they do, they never really get around to it anyone, because the current setup works for them.
Napster, and its near equals like Scour, all have sharing set up by default, and they both encourage you to stay online even if you're not using the program. Yesterday, on Scour, a whole bunch of people figured out that I had some somewhat rare anime videos. Instead of logging out when I was done, I just sent people messages that I wouldn't be there to monitor the transfers, and went to sleep. I think this happens more often than people think.
People love sharing; it makes them feel generous. However, it CAN'T be difficult to do. In Gnutella, it is.
-Rainbowfyre
Vericon is coming!
P2P has just scratched the surface. To say it is dead before it even gets out of the starting gate is a level of eagerness that surpasses morbidity.
There are constraining factors on p2p, but these will actually fade away as more people get broadband. Sharing will become more prevalent when it is made easy and has an obvious level of security (like Napster, where you choose which folder other's get access to). Also, as soon as it is decided what can and cannot be shared, that will open things up. I know I get a bit leary when I see people downloading my Juice Newton tunes, wondering if it is actually Juice's lawyers gearing up to sue me.
P2P may not be the next killer app, but it will become a mainstay of the internet like ftp. So let's stop paying attention to doomsayers who are just trying to be seen as prophets of the internet through Kassandra-like proclamations.
This is not the way to build a lasting empire.
Who's "many"?
Sounds like a biz journalist looking for the Next Big B2B Thing and, coming up empty, bitching about it. Last time I checked Napster was still going strong. If 30M users aren't "the mainstream masses," um, who is?
sulli
RTFJ.
One of the big (well, big to me anyway) issues about P2P networking is the number of folks like me whose primary access to the 'net is through their corporate firewall. P2P software is not nearly as useful to me, and others like me, because we plain and simple can't use it -- our firewalls come up hard against pretty much any ports other than those used for "business". Right now, I can use HTTP, FTP Get, and that's about it. And I can well imagine a day when even those will go away.
Okay, enough of the analogy! The point is that long distance bandwidth (influence) is limited. However, short distance bandwidth to a limited number of peers is not a limiting factor. So, peers only need to look in their local "neigborhood". Since each "peer" has a slightly different "neighborhood",
drum roll pleese....
the information on the P2P network will propogate reguardless of bandwith restrictions on long-range connections.
Obvious to anyone that understands how news servers work, but aparently not CNN.
science is a religion
1. They are lazy
2. They are afraid.
The penalty for possessing copyrighted material is much less than that for distributing it.
--
--
You are a fucking moron.
Of course; after all, most people can't act as a server, even if they have broadband or DSL. That's why most of the ISPs which use those two use asynchronous connections (upload much slower than download). That way, users are driven away from acting as any kind of server, but are more than happy to download files and connect to multiplayer games as a client.
From what I've heard, though, Covad uses restricted SDSL. That's nice, however, it's hard to find a reliable connection to that over here in Verizonland. I've tried to run a Q3 server on my 640/90kbps dn/up DSL connection; it wasn't pretty. My friend kept getting booted off for no reason, and pings were upwards of 300 for the clients.
"Ancillary does not mean you get to rule the world." --U.S. Circuit Judge Harry Edwards, speaking to the FCC's lawyer
However, the whole idea that P2P is at all different than server-to-server is ridiculous. Just TRY to set up a P2P connection on the net without going through an ISP.. If you can, then you ARE and ISP. You are a 'server' - whether you have clients of not is irrelevant. Even major corporations today have to go through an ISP for their connection to the backbones. My little workstation has to make just as many hops to get to Mae West as Sony's data center.
There is no technical difference between gnutella and a couple of buddies running anonymous FTP servers on their home machines. There is no technical difference between that and IRC - except for volume of bits. Bits is bits is bits. The difference, the ONLY difference, is that there isn't a corporation extracting an additional toll on the data that's transmitted. There lies the 'problem' with P2P.
If Guntella and Napster were used to share vacation photos NOBODY would care. ISP's might jack up their rates based on how much pipe you use, but that's it. If the data transfered wasn't (arguably) someone's 'intellectual property', this would not even be an issue.
People have been running private FTP servers in a P2P fashion since before the WWW made server-to-server the defacto mode of operation. Before ISP's got on the band-wagon, is was all workstation to workstation, account to account, peer-to-peer.
Just because some kid slapped a web interface onto a hack of anonymous FTP doesn't suddenly make it a different technology. Just because he made it distributed doesn't make it anything more than simply 'convenient'. Searchable FTP has existed for a long time, also since before the www. Anyone remember the Archie tool? Indexing, and making it transparent is the next obvious step, not some revolutionary break-through.
P2P is nothing new, and it is nothing 'different' than what has always been done. Servers talk to each other as 'peers' too, don't they?
Just because a bunch of corporate-types label the same technology in two different ways, depending on wether they get a cut of the profits or not, does not make one way doomed and the other saved. Just because somone calls this 'piracy' and that 'a stable business model' does not make the two ways into different technologies.
P2P, S2S, B2B... It's all the same technology. It's the same protocols and algorithms. It's all the same bits. The difference is only in who is in CONTROL of THE DATA. He who controls the INFORMATION, controls the Universe.
As for P2P 'failing' due to low bandwidth at the 'local loop', well, that's just a hot, steaming pile of BS. Ye Olde Bulletin Board Systems (the ORIGINAL P2P networks) thrived on 2400 baud.. They thrived even more on 9600, then, when 14k came, the Internet had started to mature and began to offer more 'value', farther reach and more neat stuff. But the BBS's didn't 'fail'. Not due to poor performance or inequitable sharing of files within the communities they supported. In fact, the only times BBS's were put out of business (except for their owners personal choice) it was due to... (drum roll) PIRACY and kiddie porn.
The REAL jabber has the /. user id: 13196
The REAL jabber has the user id: 13196
What you do today will cost you a day of your life
It's one that people have been making about gnutella for a while (where there is in fact a lot of overhead), but gnutella's hardly the best example.
If you haven't looked at freenet yet, then do so. Not only is it peer-peer, but it's anonymous, and it's working TODAY. There are smart folks developing it, and they're being very careful not to make the same mistakes gnutella did.
-- Anne Marie
We are hearing journalists over and over pick up on a technical truth -- the fact that p2p networks such as Gnutella are slow because their design limits practical size, most people are taking but not giving, etc. -- and saying "look! it doesn't work!" Meanwhile, the people who designed these networks in the first place (and many others) are busy finding ways to make the technology work. That's the spirit of this thing -- "Napster is a legal problem? Ok, we'll make a decentralized network. The new network protocol is slow? OK, here's one that works..."
We are just in the time between the identification of the problem and the solution. Expect to see this one figured out.
-----------
-----------
If you ever drop your keys into a river of molten lava, forget 'em, because man, they're gone. -- Jack