Lets chat about P2P some more
I'm obviously not a huge P2P fan. Gordon Haff did a far better job than I explaining some reasons why. I think there. are valid applications for P2P on private networks, but nothing on the Internet that I think is worth surviving.
My position has nothing to do with Piracy. I think the MPAA and RIAA efforts towards piracy are a joke. They spend more money and waste more government resources than should be allowed. If they spent that money and time promoting why people should go to the movies and the value of owning music, those industries would benefit far more than anything they lose to piracy.
My position is not "if it uses bandwidth, its a bad thing". Flickr, Google Video, any host that pays for their bandwidth is all right by me. If they want to give it away, go for it. I actually think Google Video is a far better solution for audio and video distribution than any P2P solution. Google is willing to subsidize the worlds bandwidth for multimedia, why doesn't everyone take them up on their offer ? Go for it Google.
My position is not related to the Internet backbone. There is plenty of bandwidth there and will be for the short and as long a term as I can envision.
My position is related to the last mile. P2P is so incredibly inefficient. You send and receive the same bytes , which means for the portion of the file you are a seed for, you are at least 50pct inefficient. The more often you supply the bytes on your PC to others, the more you impose on the network. If there is a failure somewhere in the chain of delivery and assembly on the destination device , the error recovery process makes things far less efficient. All consuming more and more last mile bandwidth. The bandwidth that defines how fast my internet connection is.
I think the position that "you pay for the bandwidth, so you can use it any way you want" isn't reality and very flawed when it comes to P2P.
P2P "works" because those who install the clients are wiling to barter some of their bandwidth in exchange for getting a file that represents something of value to them. The bandwidth obviously has significant value to the person or company asking you to contribute it. That's why torrent clients and almost every P2P client requires you to contribute bandwidth in order to receive the goodies you want.
Bottom line, you are re-selling bandwidth. For those of you who like the buffet analogy, that's like saying you paid for the buffet, so its OK to take as much jello and mac and cheese as you can carry and walk outside the restaurant and sell it or trade it. Bizarre example, but it makes the point. Just because something is not metered and seemingly not suffering from any level of scarcity doesn't mean it isn't limited in availability and costly.
Because if it wasn't costly, we all would already have 1gbs to our home via fiber or free wireless everywhere.
The reality of our bandwidth to the home scenario today is that there isn't enough bandwidth to cure all ills. Last Mile Bandwidth is constrained and expensive to grow in multiples of what we all are ready and happily able to consume with legit applications
I personally don't want to see my connections slow down so P2P users can resell bandwidth to someone who isn't willing to pay for bandwidth in order to distribute their bandwidth consuming files.
but hey, that's me.
Recent Posts
- Beating Google ? (5/14/2008)
Reader Comments
(Page 1)3. Mark I think you're wrong on this. And how can I prove it?
Akami. The primary company behind content distribution. Guess what their CDN does behind the scenes? Something that can be considered very P2P'ish. So P2P isn't bad, it's just changes the economic game.
What does that mean? What you're right about is that companies that use P2P to distribute their paid content are leveraging their consumers to distribute the content to other consumers. Is this a bad thing? I pay for my bandwidth. I don't just pay for a downstream link, but also an upstream link (I want to be able to RDP into my Windows machines and SSH into my linux boxes that I have at home). If it's cheaper to distribute content via P2P, those cost savings can be passed onto me (not saying they are, but they can be). Though it's not always cost savings, it could be time savings (say, buying a game from Valve off of steam instead of buying a retail box copy in the store). I give something to the vendor as I get something out of it, a pure economic transaction.
Efficiency has nothing to do with it. Yes, an http download is more "efficient" than an a bittorrent download in measured in "tota bytes transffered". However, its also less efficient by a different metric, namely the "use it or lose it" metric. If I'm not uploading, that bandwidth is wasted. I can upload and download at the same time w/o really impacting either one tremendously. Therefore, in a P2P system no matter how much bandwidth the content provider throws at the problem, a P2P system will always be able to have more, without really impacting the efficiency of the download. In fact, it can make the download happen faster especially in the case of flash crowds.
But as I said, it's an economic problem, not a technical one. If demand for bandwidth exceeds, supply prices will rise. If prices rise, the economic decisions that I and the content provider make will change. However, as ISPs are increasing bandwidth caps across the board without raising prices (time warner just increased download rates to 10Mbps in NYC and elsewhere, while keeping the same prices that just a couple of years ago got us 5Mbps) it would appear supply in the last mile is more available than you would think.
Posted at 1:23AM on Nov 21st 2007 by spotter
4. I'm not following how bandwidth I'm paying for is being "resold". It seens to me that you are labeling downstream bandwidth as "ours" and upstream as "theirs". When I pay my internet bill, I'm paying for a download AND and upload bandwidth. It's listed in my package plan clear as day, no confusion. If I download a torrent, I'm using my downstream bandwidth. If I seed that torrent, I'm using my upstream bandwidth. No resale, just using what has been paid for. In many cases for me it's not even a matter of "using all I've paid for". My uploads rarely exceed my total allotted "speed". Come to think of it you may be confusing "speed" and "usage" as well.
I think you need to stop seeing the internet as a tangible product, and take it for what it is. They call it the world wide web because it is just that, a web of information going to and fro, here and there. When I buy access to the internet I'm buying access to that web in whatever direction or form I choose, and is allowed by my terms of service.
That "last mile" you write about is outdated information. The "last mile" issue is constantly being addressed through ongoing infrastructure upgrades. Comcast has also increased its bandwidth "speeds" twice in the last few years in many of its markets. There hasn't really been a bottleneck of any kind in quite some time, and the current bottleneck hotspot is dealing with the cable spectrum, not the the internet.
Posted at 1:43AM on Nov 21st 2007 by Tony
5. Mark,
What was the aha moment when you decided that p2p was not such a great idea? You put your money on the other side of the argument when you invested in Red Swoosh, and it looks like you came out of that deal just fine since Akamai bought them for $15MM.
I don't buy your argument that p2p makes the last mile inefficient just because clients are sending and receiving the same bytes. I do agree with the inefficiency due to the retransmit.
p2p will actually make the last mile more efficient once critical mass is achieved. If a host does not need to send a file more than one router hop away, then it is more efficient to have the content stored in the last mile. If there were 10 people on your block, and they all wanted a certain piece of content, then that would be 10 streams from Google video from their closest distribution node. The same bytes would get transmitted 10 times from that node to the last mile and then to the end user. If the content was cached locally and served up via p2p, then it would all get transmitted on the network segment serving the last mile and not touch the upstream network. This in turn will free up the upstream part of the network to provide better service to the last mile.
Posted at 1:47AM on Nov 21st 2007 by Doug
6. Flickr are known censors.
YouTube has terms of service with many elements based on the United States puritan background and modern-day religious fundamentalism.
Both are owned by the two most visible gatekeepers on the Internet. Oops, did I say "gatekeepers"? I meant "enablers" ;-)
It is so seriously hard to understand why you appear to be suggesting that the answer the last mile issues and bit-waste is to oblige the gatekeepers.
Why is it extremely hard to setup a distribution network for media? Isn't it in large part because the cost of entry is so high? If the very few gatekeepers can make such extreme amounts of income, wouldn't the barriers to entry just be growing larger?
I realize it's all complex, we all do. But P2P, not blogs is the most powerful mechanism for free-flow of information in the world today. It tears down just about every known barrier to distribution and communication in the world today.
Oh, and about the "gatekeepers" term: If you can't be found on Google, do you exist? Google is a massive reputation network with a secret formula for determining whether you are on page 1 or 1,000,000,1 of their search results. It's of course extremely useful, but its still not transparent, nor is reputation the only consideration in the pursuit of personal freedom.
P.S. If Comcast is one of the few networks willing to carry HDNET, isn't it good form to mention your conflict of interest along with your endorsement of their practices?
Posted at 1:54AM on Nov 21st 2007 by Marcus
7. Mark,
I'm paying for 1.5down/1.5up. Thats mine to use how I please.
To deal with this oversubscription issue: I understand part of the ISP model is to oversubscribe. I'm ok with that, but I sure as hell better be getting my 1.5/1.5. If not, you need to either make me pay more to really get my 1.5/1.5 or fix the oversubscription ratio so that you can deliver the class of service that you are selling.
How can you agree with the idiots that say you can't do what you what with the bandwidth you pay for? Thats like selling an unlimited bus pass that will only take you certain places. Or how about if you paid for a certain amount of electricity, but since the power company oversubscribed your neighborhood they had to drop the power output to your homes during busy times?
Come on, this is nonsense. The cable companies are ripping us off by pulling stunts like this to make their over-over-subcription model work. I'm surprised to see someone as brilliant and forward-thinking as you jump onboard with these clowns.
8. Mark,
You do realize that P2P is used for a LOT more than just file exchange & bittorrent?
The largest example is Skype. Skype is built on a core of P2P. Rather than a central switching system, Skype is a large, p2p grid. And as a result offers an amazing service to millions of people (literally changing the world in many places - and skype is not the only voip system that uses p2p though it is probably the best known and most successful)
Joost, which is much more questionable as to whether it will succeed or not is also built on a core of p2p.
The internet IS designed to be p2p - from the routers to servers to core functions like mail routing they were built to have peers exchanging data with each other. What most people don't realize (but which is what p2p at its most basic is doing) is that modern personal computers ARE capable of also being servers - i.e. not just requesting data DOWN from the network but also sending it back up to the network.
And the network is built for this - but in many places with some built-in chokepoints. Most DSL in the US, for example, is asynchronous - with faster download speeds than upload. This is a decision made by the telcos as to the technology they would deploy and what tradeoffs they chose to make (distance over speed, number of systems they had to install in their backoffices etc).
But, prior to widespread adoption of p2p type technologies most users on the internet did not make much use at all of their upstream capabilities. It was there, and used lightly when entering data on a webpage, sending an email etc.
I would also note that P2P works BETTER on a large, widely distributed network than on a small, private network. On a widely distributed network any single transaction is NOT between one computer and a single other computer (i.e. replicating the traditional client/server model) but rather is one computer to many computers - each providing a piece of the data needed. This results is some overhead to coordinate, but also allows for a great deal of flexibility, redundancy, reliability, and speed. Any single machine and that machine's connection may be close to saturated - but is usually not the only source for a key bit. In the case of a torrent, a well seeded torrent downloads very fast - and as more people request it the speeds go UP (more peers) not down.
i.e. p2p grids get better as more parties participate in them (if the coordination code is well written) NOT as so much else on the network worse with scale.
As others noted, content delivery networks are, in essense, p2p grids - but where most (or all) of the nodes are controlled by a single entity - but they are still usually distributed across the larger Internet and generally use p2p like technologies to spread out the requests (and loads) across multiple machines.
Google itself is the operator of what most people would agree is the largest grid out there (or if not among the largest and probably the largest owned entirely by one company). Their search index and other apps are spread across 1000's of computers, distributed globally, and their response to a request is then distributed likewise.
So though I think i get what you are talking about - I have to disagree.
That's not to say that better technologies or restrictions on poorly designed systems shouldn't be enforced. Nor that bandwidth providers can shape their traffic to a degree (though I'd rather they concentrated on getting true broadband to US homes - i.e. fiber to the home, and delivered in a synchronous no asynchronous manner)
Shannon
Posted at 2:47AM on Nov 21st 2007 by Shannon Clark
9. Mark, There are legitimate P2P applications that I don't think you want to lump in with anonymous file sharing (or distribution). Consider screen sharing, either one-to-one or one-to-few. Consider VOIP and voice conferences (Skype, iChat). There are also servers located in the home that serve up data to families when they are away, etc. You're rightly concerned about upstream bandwidth on networks you share with your neighbors. But it's important to distinguish between services that legitimately use that bandwidth and those that might not.
Posted at 2:49AM on Nov 21st 2007 by Brad Hutchings
10. Mark,
Just a quick thing to add. I think the buffet analogy needs fixing.
It's OK to take as much jello ... 2048 grams, for example, for arguments sake
But I'm also paying the buffet to give away 512 grams of it however I please.
I give away all 512 grams of it, and allows another buffet go-er to give me 1024 grams of cheese, because he wants jello, and the line is too long.
Sounds weird, but it does make sense. I own my download, and I own my upload. It's not reselling. If it was, wouldn't uploading videos onto Google Video, and streaming from Google Video be the same thing? We are taking and giving bandwidth, and if no one uploads, it would fail... just like P2P.
If its the last mile, blame the ISP for advertising what they cannot offer.
Posted at 3:16AM on Nov 21st 2007 by Derek Wong
11. Mark,
Let's say I've discovered something rare and I want to share with my friends. 1GB DB of fantasy football information, telescope shots, video of a friend's birthday. And since Google Video is not truly private, there is a reason I would not want to go there, let alone current Flash video quality.
If I HTTP POST these things to my friends(let's say 50), I'm sending the same bytes 50 times - and being the initial seed, I never downloaded, so I'm woefully inefficient. Even downloading my discovery initially puts me at what, 2% efficient? Much, much worse than your 50% above.
I can't remember the math as of right now, but I believe that with BitTorrent, the more popular a file becomes, the more likely that the actual efficiency will approach 50%. 50 People might get me 30%.
"I personally don't want to see my connections slow down so P2P users can resell bandwidth to someone who isn't willing to pay"
Is that someone an individual (like me) or a corporation? As far as I'm concerned, I am paying Comcast a set fee for a connection, as much as any hosted service has. A corporation relying on me - that is something else entirely, and you haven't distinguished that in your posts. Comcast hasn't in their actions, either.
That Comcast has set up their network in expectation being in an ADSL world while we are rapidly transitioning to an SDSL one is a failure on their part. Well, DOCSIS as a whole you could say.
Here's something else - if I've paid for Comcast's business cable modem, should they still be blocking my P2P apps? Since this will still slow down your last mile.
Posted at 4:39AM on Nov 21st 2007 by Dave
12. Hey Mark,
Just one question. Were Travis and the redswoosh p2p guys doing bad things in your mind when you invested, and before they sold to Akamai?
Posted at 5:54AM on Nov 21st 2007 by Hank Williams
13. The problem is that while bandwidth has a cost, it's being given away for more or less for free; when users are getting theoretically unlimited bandwidth, the incremental cost for more of it (say, for a P2P application) is zero. So if you can use something that costs you nothing to get something you want, that's an economically rational choice.
The problem with the approach providers like Comcast are taking is that it's way too indiscrimate; there are examples of them blocking things like people synching their Lotus Notes databases to their work servers, a totally legitimate use.
Rather than try to play bandwidth cop and shutting down the wrong people, providers need to change the economic basis that makes heavy P2P use an appealing choice. Cap the bandwidth so the incremental cost is no longer zero; people will either not use P2P so much, or will pay for it, which means that providers can make money off of it and manage their network resources appropriately.
Posted at 7:40AM on Nov 21st 2007 by John Whiteside
14. The problem is that while bandwidth has a cost, it's being given away for more or less for free; when users are getting theoretically unlimited bandwidth, the incremental cost for more of it (say, for a P2P application) is zero. So if you can use something that costs you nothing to get something you want, that's an economically rational choice.
The problem with the approach providers like Comcast are taking is that it's way too indiscrimate; there are examples of them blocking things like people synching their Lotus Notes databases to their work servers, a totally legitimate use.
Rather than try to play bandwidth cop and shutting down the wrong people, providers need to change the economic basis that makes heavy P2P use an appealing choice. Cap the bandwidth so the incremental cost is no longer zero; people will either not use P2P so much, or will pay for it, which means that providers can make money off of it and manage their network resources appropriately.
Posted at 7:42AM on Nov 21st 2007 by John Whiteside
15. The problem is that while bandwidth has a cost, it's being given away for more or less for free; when users are getting theoretically unlimited bandwidth, the incremental cost for more of it (say, for a P2P application) is zero. So if you can use something that costs you nothing to get something you want, that's an economically rational choice.
The problem with the approach providers like Comcast are taking is that it's way too indiscrimate; there are examples of them blocking things like people synching their Lotus Notes databases to their work servers, a totally legitimate use.
Rather than try to play bandwidth cop and shutting down the wrong people, providers need to change the economic basis that makes heavy P2P use an appealing choice. Cap the bandwidth so the incremental cost is no longer zero; people will either not use P2P so much, or will pay for it, which means that providers can make money off of it and manage their network resources appropriately.
Posted at 7:43AM on Nov 21st 2007 by John Whiteside
16. Mark, I have to agree with many people here, going so far as to say that P2P is a more efficient use of the last mile. Very few people make anywhere near full use of their upload allocation, which is what is being used most when seeding. If you are in need of preserving your upload pipe then I would really like to know what for.
I would also agree with people here that say they should get to use what they pay for. If I am paying for a 24/1 down/up connection then that is what I should be able to use. Now we all know that practically that is not true, in which case it should not be sold as such. That's less a comment to you than a rant about the poor state of reality in the consumer ISP market.
I really believe that bandwidth should be the only differentiating factor in ISP connection fees and that they need to be forced to be technically able to back what they sell.
Posted at 9:05AM on Nov 21st 2007 by Matt Large
17. It boils down to whether one thinks we should pay-per-bit for content transfer, or would prefer to pay-per-bps for bandwidth.
For example, I currently buy 8M down/768K up from Comcast. And that, by golly, is what I should be able to use, regardless of how. Comcast isn't billing me for bits, I'm paying for bandwidth, for the ability to shovel that many bits per second up or down these hoary tubes.
In contrast, my webspace provider charges me $1 per gigabyte of data transfer. And, (unlike Comcast), I fully expect I will get exactly what I pay for.
So, if you've paid for bandwidth and you're crying because some other users actually USE the bandwidth they've paid for, (which Comcast appears to have oversold by many times over what they can actually provide), then you need to review your ideas of supply and demand, and ponder over just what it is you paid Comcast for. Comcast needs to do likewise.
Posted at 9:57AM on Nov 21st 2007 by Gregory Bloom
18. Mark,
I live in a developing country, but am a big US sports fan. At times I have tried to pay my local cable operator, MLB, and NBA digital media real American $$$ to get the games that I want. The cable company can only give me about a tenth of the games that I'm looking for and MLB and NBA streaming are too fast for my connection (why don't you guys offer a lower bandwidth???). So I go to China and the good folks there will happily pirate NBA games that I want to see with the ads included. The programs are inefficient and do slow down my computer, but are far MORE EFFICIENT than any of the NBA.com or MLB.tv servers I have tried to access in the past.
P2P will not outcompete google video or youtube as a hosting mechanism, but that is b/c 90% of the traffic is pirated music and video (mostly pornography). But it absolutely destroys both MLB and the NBA in terms of streaming video delivery systems. Care to explain why?
Posted at 10:10AM on Nov 21st 2007 by dan k
19. For those of you who like the buffet analogy, that's like saying you paid for the buffet, so its OK to take as much jello and mac and cheese as you can carry and walk outside the restaurant and sell it or trade it.
This analogy falls apart. My purcahse of a buffet allows me to get an unlimited amount of food. My ISP does not let me get an ulimited amount of bandwidth. Rather, they offer me 6 Mbps downstream and 1 Mbps upstream--and I should be able to use all of that.
The problem is clear. The ISPs are overselling their capacity. If the model doesn't work for them to live up to their offer, they need to change the offer.
Posted at 10:24AM on Nov 21st 2007 by Chris Woods
20. Wow,
You certainly have picked a topic with lots of opinions :-)
Much of it seems valid to me, could you maybe elaborate on the 50pct inefficient for sending the same bytes. I didn't quite get that part. I'm sending them to different users right, not the same user over and over? Same as if I had an ftp server, I'm serving the same bytes over and over to differernt clients, but it's not 50pct inefficient. Maybe I've misunderstood though.
I personally don't think P2P is the problem. It's just a tool like a hammer or splitting an atom. I think the way the tool is used can become a problem to the extent it effects others.
Posted at 10:41AM on Nov 21st 2007 by JP Russell

1. personally, i agree with you. however, i dont think the typical user consciously cares about (or is even aware of) the cost of bandwidth. slower download = "slow computer" and may result in disrespectful putdowns shouted at computer monitor and/or erratic over-clicking of mouse button. they know not what the cause is for this reduced performance or how to improve the process.
the issue may be more with entitlement, as people feel that they should be able to have access to any and all things quickly and at no cost, as well as share property with others at no cost. whatever system is perceived by the user to meet their impulsive desires more conveniently is the vehicle used to do so, and P2P is less threatening from a "is this legal" standpoint. in addition, i would venture to say that circa 95% of internet downloaders have no clue what "P2P" even is. And if you are a host of said information and are assuming that you wouldnt lose customers to slower connections, why pay to centralize if you can P2P?
Posted at 12:50AM on Nov 21st 2007 by Seams