Follow Slashdot blog updates by subscribing to our blog RSS feed

 



Forgot your password?
typodupeerror
×

Researchers Decentralize BitTorrent 262

A Cow writes "The Tribler BitTorrent client, a project run by researchers from several European universities and Harvard, is the first to incorporate decentralized search capabilities. With Tribler, users can now find .torrent files that are hosted among other peers, instead of on a centralized site such as The Pirate Bay or Mininova. The Tribler developers have found a way to make their client work without having to rely on BitTorrent sites. Although others have tried to come up with similar solutions, such as the Cubit plugin for Vuze, Tribler is the first to understand that with decentralized BitTorrent search, there also has to be a way to moderate these decentralized torrents in order to avoid a flood of spam."
This discussion has been archived. No new comments can be posted.

Researchers Decentralize BitTorrent

Comments Filter:
  • It's a good start... (Score:5, Interesting)

    by Smidge207 ( 1278042 ) on Tuesday October 28, 2008 @02:04PM (#25544481) Journal

    ...and hopefully with this companies will start to use BT as an alternative to http/ftp. The downside is that you have to have a client, but I bet that browsers will have integrated BT support soon (the new Opera does, FF has a plugin). And the savings for the server range from a LOT to none, and even none can't hurt, since if nothing else you at least have a great download client able to resume downloads, download huge files, etc.

    =Smidge=

  • by AmiMoJo ( 196126 ) on Tuesday October 28, 2008 @02:06PM (#25544513) Homepage Journal

    BT is popular because you can go to a reputable listing site, find a well seeded and good quality torrent with comments by others to back it up and download it quickly. Compared to the chances you take searching traditional P2P systems, full of dodgy encodes, fake file names and incompletes it's obvious why people turn to BT first.

    • by MasterOfMagic ( 151058 ) on Tuesday October 28, 2008 @02:09PM (#25544583) Journal

      Yes, and you can still do this even with Tribler. They're not mutually exclusive.

      • Re: (Score:3, Interesting)

        by thepotoo ( 829391 )
        Yeah, but the interface is clumsy (pretty, but still clumsy) and there doesn't seem to be any way to read comments on the torrents. The ability to sort by seeders is nice.

        But, what I really want is a way to sort by seeders AND 4 or 5 star ratings AND filter by category AND quickly view tags (dupe, spam, nuked, wrong category, etc) on a torrent. No site lets me do this, but Demonoid comes pretty close with filters.

        Decentralization is a pretty good idea, but it's certainly a long way from being ready for pr

    • by imroy ( 755 )

      The eMule network has ed2k URI's [wikipedia.org] which use a hash function to identify a specific file. I don't see them used much, but I'm not a big P2P'er.

      • I use eMule quite a bit and the URI system is great. All the ed2k sites use it.

        sharethefiles.com and tvu.org.ru are two big ones - they do for emule what piratebay does for torrents (just using those URIs in place of .torrent files).

      • The eMule network has ed2k URI's [wikipedia.org] which use a hash function to identify a specific file. I don't see them used much, but I'm not a big P2P'er.

        I didn't get from TFW whether the ed2k 'trackers' still point to servers with the files. What are now bittorrent tracker sites would be better off simply being attestation databases based on hashes (This hash represents X and Y people rate it Z), and letting the clients find the files themselves. Some sort of distributed hierarchical database probabl

    • It's still a single point of failure.

  • by bconway ( 63464 ) on Tuesday October 28, 2008 @02:10PM (#25544609) Homepage

    Full decentralized, search capabilities, with many people able to share pieces of the same file... I think we already have something like that [wikipedia.org].

    News flash: Centralisation is a strength of BitTorrent.

    • News Flash: New feature does not remove older.

      I just don't see how this would hurt torrents in any way. (unless you talk about cliques of people worried about their private trackers, who might be exposed to public by some seach feature, in which case solution is not use client that does that.)

      Also, Torrent was hardly centralized lately, popular cients support several tracers per torrent, user-to-user peer exchange and tracer discovery. Hell, those "centralized servers" usually just offered torrents with tra

      • by Goaway ( 82658 )

        I just don't see how this would hurt torrents in any way.

        It wouldn't, and nobody said it would. The point is just that nobody wants this.

        • Re: (Score:3, Insightful)

          I do. Having non-tracker searching might resurrect some of my current dead torrents. I hate being stuck at 80% forever, and there might be a seed somewhere out there just waiting to be discovered.

      • But BT search _is_ centralized, and search is what this article is talking about. Downloading material is much less of a crapshoot if you can reliably know who uploaded a file and see what other people say about it. I don't relish the idea of having to sift through pages of unreliable search results again a la gnutella; I'd much rather stick with what's in place.

        The decentralization you're talking about all happens after you've chosen what file to download. And it's all predicated on a central tracker, anyw

    • News flash: Centralisation is a strength of BitTorrent.

      What makes you think centralisation is a strength? I liked gnutella just much better for actually finding stuff, except that BT is faster for downloading.

      Newsflash: with a decent client that let you ban junk, and enough common sense to see the junk patterns, gnutella was (and still is) very reliable.

    • Re: (Score:3, Insightful)

      by TubeSteak ( 669689 )

      News flash: Centralisation is a strength of BitTorrent.

      Ever since the demise of Napster and Grokster, followed by the public death of eDonkey, several large ED2K servers and various gnutella clients centralization has not been a strength and everyone has been running as fast as possible in the other direction.

      The problem with centralization is that it gives the **AA a big fat target to aim their lawsuits at. And they've quite successfully sued P2P companies into oblivion and hassled websites out of existence.

  • by Anonymous Coward on Tuesday October 28, 2008 @02:15PM (#25544691)

    According to the website, Tribler will exchange torrent downloading history by default.

    • Re: (Score:3, Informative)

      Yeah, and it also states it can be easily turned off by "disabling the recommender in the Preference menu".
      • Re: (Score:3, Interesting)

        by TubeSteak ( 669689 )

        Yeah, and it also states it can be easily turned off by "disabling the recommender in the Preference menu".

        I think you're missing the point.

        It's a stupid and arguably dangerous default setting to have enabled. The only way the company could make the **AA's lawsuit campaign any easier is to e-mail them the name of every torrent you download.

  • Researchers plans (Score:2, Insightful)

    by philspear ( 1142299 )

    1. Decentralize bittorrent
    2. Share pirated stuff
    3. ???
    4. profit
    5. Cure cancer?

  • by imroy ( 755 ) <imroykun@gmail.com> on Tuesday October 28, 2008 @02:29PM (#25544913) Homepage Journal

    Is it just me, or is the BitTorrent world slowly converging on features and an architecture that the eDonkey network [wikipedia.org] has had for years?.

    I mean, BitTorrent started out as a way to download big files, like Linux ISO's. Then people started making big torrent search web sites, similar to eDonkey servers. Then people made BitTorrent clients that had a queue of downloads (e.g utorrent), quite similar to eDonkey clients. Now these people have made Torrent searching distributed, just like eDonkey and Kademlia.

    I've never been much impressed by BitTorrent (gee, can you tell?). Just what is it that makes it more popular than eDonkey/eMule? Is it just the reputation and hype that has built up around "Torrents"?

    • by fmoliveira ( 979051 ) on Tuesday October 28, 2008 @02:39PM (#25545077)
      with ed2k I enter in the 2000th position in 2000 different queues. with torrent my download starts almost immediatly at the top speed my connection supports. I don't like the work of emule developers at the protocol, and they aren't very receptive of suggestions. I think the users voted with their downloads.
      • Re: (Score:3, Interesting)

        by imroy ( 755 )

        with ed2k I enter in the 2000th position in 2000 different queues.

        Ok, I'll give you that. It often takes a while to start a download, especially if it's not widely available. eDonkey seems to be setup for college students - run it 24/7 and everything is queued. I wonder how BitTorrent does it differently. Surely not every BT download starts immediately - there can't always be enough idle peers (with the content you want) to make that possible. Is your experience mainly with new or old content, or both?

        • by Endo13 ( 1000782 )

          Bit torrent starts to upload literally as soon as you have data that any other peer needs. It seems to me like new peers are given some data ASAP for that very reason. And then of course, as soon as you have some data to upload, another peer/seed is relieved from uploading that particular data, giving it more upload cap to send you more data, etc. As long as people aren't too stingy with their upload it works pretty well.

        • by Carnildo ( 712617 ) on Tuesday October 28, 2008 @03:29PM (#25545865) Homepage Journal

          You seem to have some misconceptions on how BitTorrent works. Basically, when you start a torrent download, your client asks the tracker (a central server that's keeping track of things) which computers have the download in question. Your client then asks those computers for pieces of the whole download. The pieces come in random order, and it might take a while for you to get the whole file, but the strength of BitTorrent is that, by asking many computers for small pieces of the file, you're getting a share of the collective upload bandwidth of every computer that's got part of the file, rather than getting the complete upload bandwidth of a single computer. This lets the download start immediately, and means that even peers that don't have the complete download yet can help speed things up for you.

          • by funfail ( 970288 ) on Tuesday October 28, 2008 @03:54PM (#25546223) Homepage

            And that's different from eMule/eDonkey protocol how?

            • Re: (Score:3, Interesting)

              I think the difference is simply in the number of files and the time period.

              torrent: open for only a few hours and for a single download. Usually for recent tv shows.

              amule: on 24/7 and am sharing out all of my anime fansubs(2000-ish individual files). Some of them over a decade old.

              My overall upload bandwidth is throttled to about the same level for each, but the upload bandwidth for ed2k is distributed amongst my entire collection of files. Therefore people have to wait a lot longer to get a particular

          • The pieces come in random order

            Actually the client is free to choose the order of the pieces, which is why different cleints have different real world performances. A common trick is to first ask for anything and everything then once your d/l at full speed start looking to get the rarer pieces, which ofc leaves the end of the download getting common pieces so even though there are few of them it will not decrease the d/l speed.

        • by Mr2001 ( 90979 ) on Tuesday October 28, 2008 @05:37PM (#25547791) Homepage Journal

          I've only used eMule, so I don't know how much these problems affect other eDonkey clients, but in my experience these are the big issues:

          1. You can only start sharing once you've downloaded an entire piece of the file. The same is true of BT, but eMule pieces are big and have a fixed size (around 9 MB). Torrent piece sizes are variable, and they're often less than 1 MB. This means you can start sharing sooner, especially since...

          2. ...eMule severely limits the upload speed per connection. If you set your upload rate to 30 KB/sec, you'll end up with 10 connections, each uploading at 3 KB/sec. At that rate, it takes nearly an hour to transfer an entire piece of the file, and until that's finished, the peer can't share any of the data you've been sending him.

          3. eMule's credit system is mostly only useful when you're downloading a group of files that are shared by the same users who are also interested in some similar files you have (i.e. you share S1E1 and gain credits that you redeem when downloading S1E2). BT provides immediate gratification: your uploads are almost always reciprocated right then and there.

      • with ed2k I enter in the 2000th position in 2000 different queues

        You're exaggerating ridiculously, a file with 2000 sources you should start getting within a couple minutes with emule. A file with only a few sources may have large queues that take a while.

        But thats part of the design of ed2k vs BT. BT is designed for single downloads vs ed2k as batch. With BT if you're seeding 3 different torrents, you have 3 different "queues". With ed2k, when you get a queued 2000 that's probably because there's only a few sources sharing 10k+ files and the file wouldn't be present

        • Re: (Score:3, Interesting)

          by Hatta ( 162192 )

          You're exaggerating ridiculously, a file with 2000 sources you should start getting within a couple minutes with emule. A file with only a few sources may have large queues that take a while.

          Having queues at all is the problem. Don't make people wait to participate in the swarm. Give them data immediately, so they can upload it to others.

    • by BPPG ( 1181851 )

      I've never been much impressed by BitTorrent (gee, can you tell?). Just what is it that makes it more popular than eDonkey/eMule? Is it just the reputation and hype that has built up around "Torrents"?

      I've kind of wondered this as well. I'm not a heavy bittorrent user, but it is what I mostly use for P2P.

      The only reason I can think of is that it makes sense on a browser-centric desktop, you learn about it and use it with your browser, whether it's IE or firefox or lynx.

    • by Timmmm ( 636430 ) on Tuesday October 28, 2008 @02:50PM (#25545231)

      It's the fact that

      a) The torrent sites are easy to search, have good files and few fakes.
      b) The tit-for-tat algorithm does a pretty good job of ensuring people upload stuff to you. Every other P2P software I used before bittorrent was slow and unreliable.

    • I've noticed downloads in eMule files containing .torrent files and wondered why. I combined my shared directories between all my sharing programs into one directory. Then it dawned on me. When I downloaded something in a torrent where the download contained the .torrent file, eMule would share that .torrent file. So someone searching in eMule (or Limewire) could come along and fine my shared .torrent file as a source to the shared file. By the same token, I can search in eMule for a .Torrent, download

    • Re: (Score:3, Interesting)

      by Goaway ( 82658 )

      Is it just me, or is the BitTorrent world slowly converging on features and an architecture that the eDonkey network has had for years?.

      No, it's not.

      What is happening is that there are people who think that what BitTorrent really needs is to become eDonkey. And then they make a big deal about how they are going to turn BitTorrent into eDonkey.

      And then they find out that pretty much nobody wants eDonkey, and that's why they are not using it, but using BitTorrent instead. And their wonderful project slowly dies and is forgotten.

      And then we wait half a year, and the next person steps up to the plate to turn BitTorrent into eDonkey.

  • So does this lead to TPB (and others) becoming largely redundant? If not right now, at least the writing is on the wall for them. I mean if you can search without TPB, then why would you need to go there, apart from maybe checking out their legal pages.

    • by Endo13 ( 1000782 )

      apart from maybe checking out their legal pages.

      But isn't that enough reason? I mean, they have some really good advice there. Like, make sure you use a good quality retractable baton.

    • by Goaway ( 82658 )

      They didn't become redundant the last couple times somebody tried to make this, so I doubt they will become redundant now.

    • TPB is also a bittorrent tracker, something that users mostly ignore due to lack of knowledge of the bittorrent protocol. You can get your torrents from anywhere, it will still need to connect to a tracker that will keep track of who is sharing the pieces that you want. Often people say that "lol piratebay sucks, use bigger sites like isohunt", but then fail to see that most of the isohunt torrents are using TPB as a tracker.

      The role of the tracker is somewhat going away with DHT and PXE, but is still very

    • Because the interface is pretty good, the comments give good feedback and you agree with their ideals (sweden is not in US jurisdiction). personally i find sites like mininova with its RSS feeds much more useful than limewire ever was for finding stuff to download.

  • Stupid Question (Score:5, Interesting)

    by bendodge ( 998616 ) <bendodge@bsgproY ... s.com minus poet> on Tuesday October 28, 2008 @03:18PM (#25545721) Homepage Journal

    I know this is a naive question, but how does a client find any peers to query without a centralized server to get a list from?

    • DHT

      http://en.wikipedia.org/wiki/Distributed_hash_table [wikipedia.org]

      Basically they find matches of the same BITTORRENT files, not the same FILE, and peers and seeds are connected by the hash of the bittorrent file to each other if and only if they match bittorrent hashes.

      tl;dr they connect to each other only by matching bittorrent files using a magic server.

      • Almost right. The GP is asking how to connect to the overlay network (in DHT terms) to find peers to query, rather that how to find the peers that match a query, which is where a DHT comes into it.

        Without delving into the Tribble source I would guess the other reply below is correct that a hosts.txt lists some long-lived peers in the network to bootstrap the search from. Once one of them has been contacted then more local peers can be found, and then the DHT protocol can be run over the top.

        In fact I didn't

    • I know this is a naive question, but how does a client find any peers to query without a centralized server to get a list from?

      They query 224.0.0.1 :)

    • Of course I haven't read the article, but one might suppose there's a hosts.txt along with the client. Whenever a new connection is made, the parties remember each other and add to the hosts.txt. There's some kind of sub-protocol that lets parties exchange peer information.

      Probably some fancy math can show you that if you connect to 2, or 5, or log(n), or 1/(2^n)! random peers, you can reach everybody in your connected component with a ttl of O(f(n)) for some f, and your connected component will split wit

  • Tribler and Cubit (Score:5, Informative)

    by BernardWong ( 927564 ) on Tuesday October 28, 2008 @04:22PM (#25546675)

    Disclaimer: I'm one of the authors of Cubit

    Tribler takes an interesting approach to the distributed search problem -- collect Torrents in the background and perform on-demand searches locally. To improve recall, skew the Torrent collection to collect mostly from those that have similar interests.

    It does raise a few questions. Search quality for less popular Torrents will likely be affected. Searching for Torrents outside your typical interests may also be problematic. And given a Torrent may in theory be replicated to every Tribler client, there is some bandwidth concerns.

    I guess only time will tell if limiting search to only the files that have been previously downloaded by one of your peers is sufficient for most users.

    Cubit takes a different approach -- perform efficient, distributed search over all the available Torrents in a manner that is resilient to typos and spelling mistakes (from both the search string and the content). Rely on a separate mechanism (such as user comments or a reputation system like Credence [cornell.edu]) to determine good Torrents from SPAM in the search results.

    The approaches seem complimentary, and I'm looking forward to testing out the new Tribler once the website recovers from the Slashdot-ing.

  • I still remember when suprnova closed down and they started development on eXeem which sounds to be exactly what they are trying to do here. That project didn't even last for a year.

    I believe that those researchers will fail the same way as eXeem failed.

Our OS who art in CPU, UNIX be thy name. Thy programs run, thy syscalls done, In kernel as it is in user!

Working...