Words by c.z.robertson

P2P porn statistics

2001-07-29 01:00:00 UTC

I now have some actual numbers relating to yesterday's P2P porn comments: According to Gnutellameter.com, the top ten searches on Gnutella constitute less than 2% of the total searches on the system.

term paper

Romy

2001-12-04 00:18:00 UTC

Hello. I have no idea if this is the right site to be on but I am guessing it is. Here's my problem...I am a freshmen in college and I have a term paper due in two weeks for my Philosophy class. The topic of my paper is going to be a women's view on porn and how I, as a woman, believe that porn does not degrade women. My question is this? How many WOMEN in the United States watch porn and like it? I don't know if you would have any information on that but I can't find any other sites to tell me. Thanks a bunch.

Women and porn

Tammy

2002-05-15 04:37:00 UTC

i agree... :) that's all. I like porn just as much as the next person... if not more. :)

Porn file sharing statistics for other peer to peer networks?

Karen

2002-06-25 09:57:00 UTC

Like KaZaA (or the other fasttrak networks?) or WinMX?

It seems to me like the networks that focus on larger files tend to move more porn, like http://www.peer2peerporn.com or http://www.edonkey2000.com or http://www.leechnet.com or usenet -- since they move the large files, it looks like most of the files are adult movies and complete gallery collections.

I'd like to know more about this -- does anyone track "all" the files being traded?

colin_zr - http://rtnl.org.uk

2002-06-26 22:24:00 UTC

I'm not very familiar with any of the other p2p networks, but it seems to me that gnutella is certainly more suited to transferring large files than usenet. But usenet still manages to shift a lot of porn. (I have no idea about the relative quantities of course.)

In talking about this, it might make sense to have some proper metrics so that we could have some theoretical way of quantifying what we're talking about. This was all inspired by a study which talked about the queries made on the gnutella network. I'm quite firmly of the opinion that queries are not a particularly good way of judging the uses to which a network is being put. After all, there's no guarantee that a query will return any hits. Similarly, I don't think that the files being offered are a good judge of the uses of a network. On my website I offer a whole bunch of tunes that I've written, but that doesn't mean that people are downloading them.

So it might be good to limit ourselves to talking about actual file transfers... which raises its own problems: On gnutella, while the queries and responses pass through intermediate nodes, the actual file transfer connections are made directly (using http I believe), so monitoring them will be next to impossible.

Another possibility is to monitor the intersection of queries and offered files: the results from queries. I spent a little time doing this with gtk-gnutella and it was dominated by porn and music in roughly equal measure. But still, that doesn't tell us about file transfers.

On usenet the situation is rather easier to monitor. Any ISP can (within the limits of any relevant laws) monitor the uses of their nntp server.

I don't know about the specifics of Freenet, but I think that it might be fairly easy to set up a node that could monitor the uses of the network (though it wouldn't be possible to identify the participants) since all the file transfers go through intermediate nodes.

Unfortunately, I really don't know anything about the architecture of the other p2p networks.

To be honest, I haven't really looked into this issue too deeply, but I suspect this is an area dominated by supposition and conjecture rather than hard figures.

ps. I took the liberty of removing the html from your post. My code doesn't handle it at the moment. (Actually that's sort of a lie, but I'm not going to go into the details at the moment.) I'm currently rewriting the system and I'm giving some thought to the issue of html.

you're right about searches not corresponding to downloads

karen - http://www.peer2peerporn.com

2002-06-27 13:46:00 UTC

I can't point to any published figures, but for the networks I am familiar with that can track successful downloads, the percentage of downloads that succeed (where a file or ends up on the downloader's file system) is less than 25% of all downloads attempted. This number was gathered using a modified P2P client that reported results to a centralized server.

I am not sure how to justify this figure, especially with a system like edonkey that uses multi-source-ing to retrieve their files (so a network disconnect or a system restart doesn't abort the entire transfer). The only way I can think it's happening is just a lot of people are really impatient.

Also, a good percentage of "users" are only 1-time peer to peer users. They start the app, run a few searches, usually with very generic keywords, start a few downloads, let it run for a few hours (or minutes), and then "quit and forget" and never restart the sharing app. This phenomena was also documented by the guys at Mojonation, but I don't have a link handy.

An large percentage of people leech by removing a file as soon as it completes and then quit the file sharing app. edonkey is good about sharing incomplete media, so it's hard for leechers to do that for large files on the donkey. This is not so true of the porn sharing networks -- could porn people be more willing to share? Not sure. Maybe porn people are just nice.

Usenet moves huge amounts of porn, but there will never be a centralized way of documenting the number of bytes moved or assembled into complete files...

I'm not sure if being a node in freenet would give you a big picture of the download activity -- since I think nodes will not go through the same routes to retrieve each "package", and the intermediary nodes have no idea what they're transferring and to whom it's going. With all the weight the freenet guys put on anonymity, it'd probably be the worst P2P network to try and analyze.

That said: I'm not a freenet developer, and everytime I look in to see where they're going, they're either A) exactly where they were before, or B) changed everything completely). So maybe things are different now.

Aimster (madster) and edonkey have pretty standard protocols that have been reverse engineered, and the clients will volunteer information about the files they're retrieving (and how complete the files are) if you ping them the right way. AudioGalaxy, being a centralized P2P network, probably has the most data about transfer success rates -- I know a fellow who works for them, I should hit him up for data.

Anyway, it's late and I should go to bed. It is very interesting to me though -- large chaotic networks are way cool. Thanks for fixing the HTML, I felt like a fool that I couldn't edit it.

colin_zr - http://rtnl.org.uk

2002-06-28 23:00:00 UTC

I often start downloads which I don't finish. I wouldn't say it was just impatience. It's often because it's an mp3 and I've listened to the beginning of it and decided that I don't want it. Sometimes it's because the node at the other end dropped the connection. Or it might be because I don't want to leave the computer running all night in order to get the file. And, being in the UK and having a slightly crappy dial-up account, I'm aware that I only have a certain amount of unmetered access per month.

Regarding usenet, while there won't be any centralised way to gather the data, an ISP could get a good picture of the activity of its own users. Assuming the ISP's customers where reasonably representative of the general public, you could probably extrapolate a more accurate picture of global activity than you could get for any of the decentralised p2p systems.

I think you're right about Freenet. I don't understand how it works exactly, but according to http://freenetproject.org/twiki/Main/Papers/ieee-final.pdf:

"From a node operator's point of view, the data store consists only of random GUIDs attached to opaque data."

I was assuming that it would be possible for node operators to discover the contents of their stores.

I have to say that Peer2PeerPorn looks absolutely fascinating. I'd love to know more about how it works. How do the categorisations and the collaborative filtering interact with each other?

About the html, I'm the fool, not you. This UI really sucks. No previewing, no html, the reply link on the comments is less obvious than the reply box for the main story, it's not obvious which fields are required and which aren't, it doesn't remember users... It needs fixing.

Collaborative filtering and ranking - it just might work!

karen - http://www.peer2peerporn.com

2002-07-01 16:31:00 UTC

"I have to say that Peer2PeerPorn looks absolutely fascinating. I'd love to know more about how it works. How do the categorisations and the collaborative filtering interact with each other?"

Thanks! I'm not sure how it works yet either... :-) I should put a page together describing the algorithm(s) used to try and take previous votes/impressions (weighted across a variety of categories) to rank the "value" that a piece of media probably has to the current user's active preferences.

Hopefully the current algorithms will be applicable to more than porn, but porn is the easiest way to get a large quanitity of data: porn people vote on porn, and they vote a lot: that's for sure.

We ran a version of it a while ago by linking to other people's media, and even though the original site was properly attributed (with a link), webmasters didn't like it and they had perfectly valid arguments, so it was stopped. Got a lot of data though, and that's been useful for testing new and improved algorithms.

Then the P2P idea came about while discussing the DMCA and how it could be used to help individuals (and not just the big corporations who paid for the law). So that's where effort has gone. Soon though, the ratings will be back online where everything is done on the clients. Should be an interesting summer. I'll let you know how it turns out.