Which network is best coded?

[eilveron_esp]

Which P2P network is best written in terms of speed, efficiency, not wasting bandwidth, scaling, etc?
I didn´t include Bittorent or DC because they´re too different.
And I´m really up for discussion here. G2 has been the most immpressive for me so far, but I´m not programer!

[hybridclient]

fastrack has swarming , but no partial filesahring
so this is not good for fast downloads especially for large files.

Edoneky has enourmous long queues.

So gnutella 2 is the best.

And WASTE of course.

[isus]

i think g2 is, bc it has message compression, and a lot of other useful features.

current downsides:
lack of users (less than winmx and kazaa and edonkey)
unreliable download system...

[triniti]

Originally posted by hybridclient
fastrack has swarming , but no partial filesahring
so this is not good for fast downloads especially for large files.

Edoneky has enourmous long queues.

So gnutella 2 is the best.

And WASTE of course.

Gnutella2 is not the best coded p2p stack at all. It has proven to not be scalable. It's dead, fasttrack has always been the best coded, then i would say that overnet follows. Fasttrack does need PFS but besides that it is very efective and proven scalable, G2 transfers just flat out suck. You can't beat kademlia nor fasttracks routing. G2 is based on the old gnutella routing alogorithm, perhaps minor changes and G2 downloads suck. WASTE is just IM with file transfers, it has no routing and is not really a network, it is a group of a couple of people talking directly. IMO

[CCSDUDE]

Fast Track wins hands down....

Theres a load of others with better content such as DC, but FT is a good system with the exception of Kazaa looking/workin' kind of bloated plus the fakes and the partial CRC checking results in files being fucked up over time...

G2 is no where near the best....

Edit: I'd have to say...Napster's system....even though it was central and got shut down. When I searched Napster...I searched pretty much the WHOLE user base...Kazaa may have 6 million, but you search bits of the network....with Napster I could type in some odd ball artist who never made it past 1 or 2 LP's back in the 70's and SOMEONE would have recorded it via line in somewhere along the line...

Napster fucking owned when it came to really rare audio content.

[triniti]

Originally posted by isus
current downsides:
unreliable download system...

Need anymore be said about G2? Without a decent download mechanism then there is no file sharing network. Shareaza just looks good because mike got an MFC toolkit to make it look fancy like office xp....

[random nut]

Originally posted by triniti
Shareaza just looks good because mike got an MFC toolkit to make it look fancy like office xp....

Like everything else in life, looks matter. If it doesn't look good 99% of all people are not going to try it. If Kazaa had been released today, no-one would use it because it has this ugly Win98 GUI. But maybe KMD 2.5 has the XP GUI?

[[KwestSeeker]]

the fastrack protocol does seem to have two serious flaws like: high file corruption rate and the way it identifies files (for instance when you search for a song, say something that delivers many results, it groups several versions of the same file with different lengths and bitrates, as if it was the same file so when you click on a file whose tag is a 4:03 mp3 at 256kbs you might actually be downloading something totaly different) these kinds of problems are imo way to serious to consider fastrack as the best protocol around.

[Foreverboard]

they are all good depending on what you use them for and the options you want out of your downloading program.

[random nut]

Originally posted by [KwestSeeker]
the fastrack protocol does seem to have two serious flaws like: high file corruption rate and the way it identifies files (for instance when you search for a song, say something that delivers many results, it groups several versions of the same file with different lengths and bitrates, as if it was the same file so when you click on a file whose tag is a 4:03 mp3 at 256kbs you might actually be downloading something totaly different) these kinds of problems are imo way to serious to consider fastrack as the best protocol around.

Yes, the FT netowrk is NOT reliable for file downloading. And their hashing algorithm is totally useless. It's way too easy for someone to introduce fake files and still have the file have the same MD5 hash value. The reason is simple, Kazaa/ft clients only hash the first 300K of a file, so the rest of the file is not very well protected... It's possible that the files you mentioned have the same MD5 hash (first 300K) and that could then be the reason why Kazaa groups them together. If so, then it's probably the RIAA/MPAA or whatever who's doing it.

[FileHoover]

ALL the file networks are useless when it comes to preventing file corruption.

There is no way that anyone is going to put up with hashing the file every time it has had the potential to be tampered with.

It takes several minutes to a half hour of heavy disk activity for example to hash a 700 meg movie. Any time that file is not write locked, it can be tampered with. That means if the P2P program doesn't actually keep the file open with exclusive write access it can be corrupted on purpose. That means the file cannot be moved, deleted or potentially viewed (if the viewer is coded badly to open the file requesting all access instead of just read access).

You can test your p2p program by renaming (upper case to lower case) files you are sharing while it is running. If you can do that, the files are capable of being messed with even though they are being presented as if they were verified. Be sure to check them ALL. Those that are currently being uploaded might be locked.

And don't say you can look at the write, read, create dates on the file as a tip off that the file was altered. Those can be set to any value the corruptor wants, included the values they had before he messed with them.

[eilveron_esp]

Originally posted by triniti
Need anymore be said about G2? Without a decent download mechanism then there is no file sharing network. Shareaza just looks good because mike got an MFC toolkit to make it look fancy like office xp....

Do you have any compalints about Shareaza other than it looks good?

[random nut]

Originally posted by FileHoover
ALL the file networks are useless when it comes to preventing file corruption.

There is no way that anyone is going to put up with hashing the file every time it has had the potential to be tampered with.

A user only needs to hash his files once, that's it, one measly time. No more, no less. Whenever the file is changed and detected by the p2p program, it will re-hash it. And it doesn't matter if the user manages to modify the file w/o the program detecting it. It will be detected later by another user. The automatic re-hashing is just a nice feature by the program, and not there to protect from someone tampering with files.

Whenever user A is trying to download a file from user B, user A will (when all/parts of the file is downloaded) verify that the file/chunk is correct. If it is not, then it will block user B, and possibly also notify all of user A's friend p2p'ers about user B. At least that's what a good p2p program should do. There's an algorithm (Tiger) that can do that. Probably a bunch of other hashing algorithms exist as well that can do the same thing.

[FileHoover]

Originally posted by random nut
You only need to hash a file once, that's it, one measly time. No more, no less.

Whenever user A is trying to download a file from user B, user A will (when all/parts of the file is downloaded) verify that the file/chunk is correct. If it is not, then it will block user B, and possibly also notify all of user A's friend p2p'ers about user B. At least that's what a good p2p program should do. There's an algorithm (Tiger) that can do that. Probably a bunch of other hashing algorithms exist as well that can do the same thing.
Well DUHHHH. The point of hashing is so you don't download corrupt files. If you have to download it to check if its corrupted, you missed the whole point.

[evilmegaman]

Weird you guys get unreliable downloads in gnutella2.. my downloads always finish if they have at least 3 sources. But other than g2 I like filespree's network the best. it's built very well for a udp network.

[random nut]

Originally posted by FileHoover
Well DUHHHH. The point of hashing is so you don't download corrupt files. If you have to download it to check if its corrupted, you missed the whole point.

No, the point of hashing isn't only there to check if you've downloaded corrupted files. And read the tiger algorithm/spec, or maybe it's too advanced for your "duh" brain? You don't need to download all of the file to check if the other user is sending you random data.

[shellreef]

Originally posted by FileHoover
Well DUHHHH. The point of hashing is so you don't download corrupt files. If you have to download it to check if its corrupted, you missed the whole point.
You don't have to download the entire file, only small chunks. The TigerTree algorithm is able to verify "arbitrary subranges of bytes", so you won't end up (completely) downloading a corrupt file. Read up on the THEX (http://www.open-content.net/specs/draft-jchapweske-thex-02.html) format.

[FileHoover]

Originally posted by shellreef
You don't have to download the entire file, only small chunks. The TigerTree algorithm is able to verify "arbitrary subranges of bytes", so you won't end up (completely) downloading a corrupt file. Read up on the THEX (http://www.open-content.net/specs/draft-jchapweske-thex-02.html) format.
First, tell me which p2p program uses this algorithm.

Second, which p2p program has a central database which can be used to look up EVERY file on its network to see if the hash you are downloading has been verified. A hash is just a hash. You can hash a garbage file and unless there is a central database of verified good hashes that hash is just as valid as any other.

Third, which p2p program has trusted human beings listening/watching every file and verifiying that the file is good and that its hash is good.

Fourth, "arbitary subranges" is not good enough. That's what a lot of p2p programs like Kazaa do now. It still allows corruption. Even 1% of corruption in a file is enough to ruin it.

From the Tree Hash Exchange Spec:
"Many modern peer-to-peer content delivery systems employ fixed size "block hashes" to provide a finer level of granularity in their integrity checking. This approach is still limited in the verification resolution it can attain. Additionally, all of the hash information must be retrieved from a trusted host, which can limit the scalability and reliability of the system."

[evilmegaman]

Shareaza uses tiger tree.

[ATLien]

Originally posted by random nut
No, the point of hashing isn't only there to check if you've downloaded corrupted files. And read the tiger algorithm/spec, or maybe it's too advanced for your "duh" brain? You don't need to download all of the file to check if the other user is sending you random data.

Yea, wwaaayy to advanced!

[aqlo]

I like the "tiger tree" verification process in shareaza a lot better than I did the es5 method of verifying by uploading all my files to their crappy webserver, yes I did

[FileHoover]

Originally posted by ATLien
Yea, wwaaayy to advanced!
Cool. So that means, you never get any bogus files on Shareaza?

[FileHoover]

Originally posted by aqlo
I like the "tiger tree" verification process in shareaza a lot better than I did the es5 method of verifying by uploading all my files to their crappy webserver, yes I did
Did you go to high school?

[Dividend]

No, it means you never get corrupt copies, provided the original is not corrupted. The file you download is guaranteed to be the exact file that the other persons(s) had on their computer. The internet tends to corrupt data along the way and these hashing schemes are meant to find any corrupt data that has been downloaded. SHA1 or MD5 will only tell you if part of the whole file is corrupt or not and requires the whole file to be there. TigerTree and CMD4 will tell you what part of the file is corrupt and can be checked against before the file is complete, so you can redownload the correct data before the file finishes.

Hashing does not mean that the files have been name checked by some people up above and proven to be what they say they are. It only makes sure a downloaded file is exact to the original.

[Dividend]

In any case, I'll vote for FastTrack for the fact that it's the only network proven to support over 4 million users and still search fast. I haven't used Overnet but it looks to be a very complicated network. G2 seems very promising, but it simply hasn't proven itself yet.

I also can't fathom why anybody would put ED2K as an option. It requries central servers and thererfore cannot be the best network. Although Lugdunum has done some amazing things with the servers (100,000 users connected to one computer) the network was not meant to be that way and shouldn't require anything like that to be good.

[aqlo]

Did you go to high school? Only when there was no grass, I bet you are there everyday though. Good thing too, it will help prepare you for prison.

[shellreef]

Originally posted by FileHoover
First, tell me which p2p program uses this algorithm.
Shareaza.

Second, which p2p program has a central database which can be used to look up EVERY file on its network to see if the hash you are downloading has been verified.
Shareaza. Uses the central database Bitzi (http://bitzi.com/). Not every file is listed, but that is never the case.

Third, which p2p program has trusted human beings listening/watching every file and verifiying that the file is good and that its hash is good.
Bitzi allows user commentary on each file. No trusted beings are listening/watching every file, but the mechanism is there.

Fourth, "arbitary subranges" is not good enough. That's what a lot of p2p programs like Kazaa do now. It still allows corruption. Even 1% of corruption in a file is enough to ruin it.
Kazaa hashes a subrange. TTH hashes everything, and allows arbitrary subranges to be verified. That's not what Kazaa does now. If the TigerTree hash is valid, the file can be verified against it and any corruption can be detected.

[NovaP2P]

Originally posted by hybridclient
fastrack has swarming , but no partial filesahring
so this is not good for fast downloads especially for large files.

Edoneky has enourmous long queues.

So gnutella 2 is the best.

And WASTE of course.

Impossible to make assumptions on the merits of a network when one has over 4,000,000 users at a time and the other has less than 20,000. If G2 can make it to 4 million users (which I significantly doubt with Mikes "global" searching) then id be very impressed if it was as fast as FastTrack is. Until it does, its a moot point discussing it.

[isus]

Originally posted by evilmegaman
Weird you guys get unreliable downloads in gnutella2.. my downloads always finish if they have at least 3 sources. But other than g2 I like filespree's network the best. it's built very well for a udp network.

mine always finish too (triniti)

i just think it is unreliable bc of the fact that i am on a modem, and i can go from dl'ing at 3kB/s to 0... but hey, that happens to me a helluva lot more on kazaa... so...

to what dividend said about 4 million users and it still searches fast:
you do realize that at any given time, you are connected to at MOST, 40000 people? and if you get on a shitty 'supernode', you get even less.

to whoever said napster was the best:
yes. why is napster not up here?

[crackerjacker]

hmm the best p2p network is the one that gets the job done.
sure a gui is a giving but really what works for most p2p users is getting the file.
fastrack number one
*to bad i couldnt say newsgroups* but them too
yeah napster rocked.
----------------------
and to be fair i do like the idea of dc plus because is has the least amount of bandwidth use. *this is only dependent upon the uploader and the downloader. the significent value of a p2p program is the accessibility and how user friendly it is.
---------
the future imo is subversive p2p programs. decentralized and that cant be shut down.
dc plus, gnutella and as well as waste
with waste tho it depends on the userability and the amount of trusted users on a network.
-----------
fastrack
is the best
-------------

[FileHoover]

Originally posted by Dividend
No, it means you never get corrupt copies, provided the original is not corrupted. The file you download is guaranteed to be the exact file that the other persons(s) had on their computer. The internet tends to corrupt data along the way and these hashing schemes are meant to find any corrupt data that has been downloaded. SHA1 or MD5 will only tell you if part of the whole file is corrupt or not and requires the whole file to be there. TigerTree and CMD4 will tell you what part of the file is corrupt and can be checked against before the file is complete, so you can redownload the correct data before the file finishes.

Hashing does not mean that the files have been name checked by some people up above and proven to be what they say they are. It only makes sure a downloaded file is exact to the original.

ten¡Pden¡Pcy ( P ) Pronunciation Key (tndn-s)
n. pl. ten¡Pden¡Pcies
Movement or prevailing movement in a given direction: observed the tendency of the wind; the shoreward tendency of the current.
A characteristic likelihood: fabric that has a tendency to wrinkle.
A predisposition to think, act, behave, or proceed in a particular way.

Oh please, this is bullshit. The Internet does NOT tend to corrupt data on the way. I have NEVER seen a web page with garbage in it that was caused by corruption from travelling over the Internet. I have never seen an image just randomly contain garbage from going over the Internet. If you want to believe that, go ahead, I can't stop you. The Internet would have collapsed a long time ago if it "tended" to corrupt data.

[FileHoover]

Originally posted by aqlo
Only when there was no grass, I bet you are there everyday though. Good thing too, it will help prepare you for prison. That's funny. I was telling your mother the exact same thing last night.

[FileHoover]

Originally posted by shellreef
Shareaza.


Shareaza. Uses the central database Bitzi (http://bitzi.com/). Not every file is listed, but that is never the case.


Bitzi allows user commentary on each file. No trusted beings are listening/watching every file, but the mechanism is there.


Kazaa hashes a subrange. TTH hashes everything, and allows arbitrary subranges to be verified. That's not what Kazaa does now. If the TigerTree hash is valid, the file can be verified against it and any corruption can be detected.
Well, that's a good start. If you can throw out the parts of the file that are corrupt, and get them from another user that goes a long way toward helping.

Does Bitzi hash on a low enough granularity to enable this? (throwing out the part of the file that is bad and getting it from someone else).

Is Bitzi using TTH? If so, are Shareaza users the only ones who are enabled to submit Bitzi hashes?

What prevents people from listing bogus files, with valid hashes on Bitzi?

[Evil_Dweller_01]

See I would say FastTrack...but I have something better

Ares decentralized peer to peer network!!!

Do you know why?

It works VERY similar to FastTrack yet it is being worked on everyday and it capitalizes on some of the failures of FastTrack (dont ask me which)..

It also supports pfs..and not just regular pfs
Bittorrent-like pfs

[isamoor]

Overnet all the way. Have you actually looked into those kamedlia concepts behind that network? It's pretty wild.

I just don't like metamachines. I'm just waiting for emule to bust out overnet support. That's gonna be awesome.


And I believe Shareaza isn't the only tiger-tree capable client. Most of the new gnutella clients are too. Gnutella is still being activily developed, and a lot of these neet new features are being added on. It's still not very good about large files though... too bad.

Later,

Isamoor

[thongsai]

lol file hoover.. have u even downloaded any big zip files from kazaa? im talkin about 100+ mb.. i guarantee when u try to unzip it u will get some files corrupted and u will have to download the file again for it to install correctly.. u r right that the internet doesnt corrupt ur files... but multi source downloads do.. without hashing it is pointless to get big zip files .. or data files.. but as for mp3 and avi.. u prolly wont noticed the corruption except for the ocassion blocks on some avi.. cuz the most common kazaa user's response to corruption is, " well it works for me, it plays fine.."

[CCSDUDE]

Originally posted by thongsai
lol file hoover.. have u even downloaded any big zip files from kazaa? im talkin about 100+ mb.. i guarantee when u try to unzip it u will get some files corrupted and u will have to download the file again for it to install correctly.. u r right that the internet doesnt corrupt ur files... but multi source downloads do.. without hashing it is pointless to get big zip files .. or data files.. but as for mp3 and avi.. u prolly wont noticed the corruption except for the ocassion blocks on some avi.. cuz the most common kazaa user's response to corruption is, " well it works for me, it plays fine.."

And this is why DC and BT are what I use for shit I want in a perfect unchanged format...Kazaa is like my quick fix with 192kbs stuff and sometimes 320 along with some popular movies that over 1000 users have on hand at all times....

[FileHoover]

Sure, you have to have hashing to identify unique files, but the system doesn't guarantee no bogus files. It's just a best efforts sort of thing right now

It's like Verisign authentication. Until you have some supreme trusted source that watches the whole movie and guarantees it is not munged up, or gives it an integrity rating (maybe some minor pixel blipping is ok) the ability to inject bad files into the network will still exist.

I'd like to be proven wrong, really I would, then I could incorporate the wonderful new way of guaranteeing authentic files into my code. I made some bold statements in the hopes that someone would be motivated enough to prove me wrong, and tell me why.

ZP was getting too tame. I feel ashamed that I was so bored that I participated in the the "What is your favorite word" thread. so I wanted to start a genuine argment to get some real information.

[triniti]

Originally posted by isamoor
Overnet all the way. Have you actually looked into those kamedlia concepts behind that network? It's pretty wild.

I just don't like metamachines. I'm just waiting for emule to bust out overnet support. That's gonna be awesome.


And I believe Shareaza isn't the only tiger-tree capable client. Most of the new gnutella clients are too. Gnutella is still being activily developed, and a lot of these neet new features are being added on. It's still not very good about large files though... too bad.

Later,

Isamoor
Yes Kademlia is very awesome, like I mentioned earlier. Also the GnucDNA library has tiger tree hashing for it's partial file sharing. Oh and also if you view the CVS for emule you will see they are implmenting overnet, but it is called Undernet. :fire

[crackerjacker]

I'd like to be proven wrong, really I would, then I could incorporate the wonderful new way of guaranteeing authentic files into my code. I made some bold statements in the hopes that someone would be motivated enough to prove me wrong, and tell me why.

ZP was getting too tame. I feel ashamed that I was so bored that I participated in the the "What is your favorite word" thread. so I wanted to start a genuine argment to get some real information.

---
so in essence you want to feed the fire within the ignition. why not just make your point and keep your opinion as simple an concrete as possible that way people know where you stand.

heh earth station 5 revolving around the sun platinum edition creation.
heh theres your motivation
woot woot
nuff said
btw u want to be tamed, i got a whip!
bow down to the almighty whip

hmmm
8 million times out 9 million one will million will stay strong
earth station 5 revolution
revitilize
hmm ok enuff of that
sounds like a allure commerical and stuff
-------------

[shellreef]

Originally posted by FileHoover
The Internet does NOT tend to corrupt data on the way.
I wouldn't be so sure.

I have NEVER seen a web page with garbage in it that was caused by corruption from travelling over the Internet.
You haven't, because web pages are transferred via HTTP. HTTP is intrinsically TCP. TCP is reliable; any lost or corrupted packets are retransmitted transparently.

have never seen an image just randomly contain garbage from going over the Internet.
You haven't transmitted an image over anything but TCP, then.

IP is unreliable. IP operates over unreliable carriers. It lacks error detection. Therefore, the Internet is unreliable. It may tend to corrupt data. For this reason, TCP's error detection, the system of acknowledgements and sequence numbers, is added on top of IP to achieve reliability. The Internet hasn't collapsed largely because of TCP. Have you ever used UDP? It provides no error correction, and may tend to corrupt data. Read up on it.

[shellreef]

Originally posted by FileHoover
Does Bitzi hash on a low enough granularity to enable this? (throwing out the part of the file that is bad and getting it from someone else). Is Bitzi using TTH?
Bitzi is using TTH, as well as SHA-1. A TigerTree root hash can be used to verify all child hashes; these are sent from the serving client itself and do not need to be stored on a central server because they can be easily verified. So yes, TTH hashes allow for a fine granularity.

If so, are Shareaza users the only ones who are enabled to submit Bitzi hashes?
Anyone can use BitCollider (http://bitzi.com/bitcollider/) to generate and submit hashes.

What prevents people from listing bogus files, with valid hashes on Bitzi?
Nothing. A rating system combined with a thriving community can help identify and flag the bogus files, however.

[zaphodiv]

Internet protocol version 4 uses a poor checksum algorithm on the packets
(add up all the bytes instead of a CRC). There is a significant chance of
several bit errors in a packet leading to a corrupt packet appearing to
be ok. The corruption per megabyte rate is the same for TCP and UDP

>The Internet does NOT tend to corrupt data on the way.
You are assuming that your personal experiance applys to everyone, a common mistake.
A bit error rate of one bit error in ten megabytes is not unusuall for a
modem in a pre 56K mode on a noisy phone line. IP checksums detects _most_ of those errors.
Most of the web pages you browse will be sent to you by well maintained
servers and you probably have cable/adsl.. What about when you are downloading from a load of winmodem
users with overclocked processors and dodgy RAM?

[Mel_Smiley]

I vote kazaa, Its a materpeiceofshit. It has stood the test of time as being the best performing network. I just dont use it cause I do'nt like the crowd there

[aqlo]

*to bad i couldnt say newsgroups* but them too I agree with CJ about this, usenet is a great tool for finding good downloads; but it isn't peer-to-peer.

I'm not just playing semantics here either, I have an idea. What if we distributed something that included usenet as a source within a peer-to-peer network? I don't mean just a a usenet-poster disguised as a p2p or something either, what if shareaza for example just added usenet as one of the potential networks to run and search for things and "share" (upload) to as well? You could configure it not to do anything you didn't want, but most users by default would post anything that got several search hits?

And then there's freenet, why not the same deal? Get say 10 hits on something, automatically insert it unless configured otherwise.

[eruhk]

these options suck.

probably soulseek. i use that and usenet backed up with dc++ and emule when there's absolutely no other alternative.

i hate emule with a passion.

[shellreef]

Originally posted by zaphodiv
The corruption per megabyte rate is the same for TCP and UDP
Nothing could be further from the truth.

TCP was conceived to add reliability to an unreliable, by nature, protocol. TCP insures that data arrives (acknowledgements), in order (sequence numbers), and uncorrupted. There is a possibility for error, but TCP allows for far less corruption than UDP.

[Tremaine]

#$%@ now i know why the ISO for win2k i downloaded from *guess what program* had some unreadable files, and i thought it was because i was using a cdr-w lol. But any hoo it had to re fragment my hard drive and will switch to a different OS when i eventually get a new comp.....a long time from now:blah! I learned my lesson......I will never dl a large file from kazaa, and have not since i found my new friend bittorrent.

As for the best coded network bittorrent it does exactly what it is supposed to do. As in a community i like the MP2P network which i now refuse to use with the ads and what not. Fastrack is......Overnet is too slow and all the users are mostly people who use the hybrid 210 000, G2 is just pretty, Ed2k gets the job done, Winmx ques need i say more, ES5 i still haven't tried it out what a shame, and napster was amazing but you couldn't get all the content with out switching servers but most of the content was equal between servers.

[hybridclient]

if waste would be added i would vote for waste

[Christoph]

Filetopia is encryptet,taht means for me that its the best. BUT Piolet/blubster is much faster and Pablo will add this feature too!
I would vote for Piolet!

[Lucian]

Well considering the fact that I cant really see the code for any of these, I dont really know, but the most advanced network seems to be fast track or edonkey.

[Captain_FLX]

I like the Gnutella 2 network but geez their downloads start REAL fast and go to 0.00kbps sometimes you dont even complete the files. I dont know if it's their client ? or their network who knows? Alot of people dont like FastTrack cause of the fakes I agree but if you compare both networks on WHAT you can actually download I think FastTrack is best coded.

[bernardosoares2]

triniti wrote:
Yes Kademlia is very awesome, like I mentioned earlier.

Isn't kademlia based on DHT ? I ain't no expert at all but i've read somewhere (on the GDF forums i think) that with DHT each node is responsible for storing a certain range of <key,value>, so if, at a given time, 100 000 users look for britneyspear.mp3, whose key is <whatever>, the one node responsible for the <whatever> key has to answer those 10 000 queries ?!? It may be a serious flaw for a p2p app ?

[triniti]

Originally posted by bernardosoares2
triniti wrote:


Isn't kademlia based on DHT ? I ain't no expert at all but i've read somewhere (on the GDF forums i think) that with DHT each node is responsible for storing a certain range of <key,value>, so if, at a given time, 100 000 users look for britneyspear.mp3, whose key is <whatever>, the one node responsible for the <whatever> key has to answer those 10 000 queries ?!? It may be a serious flaw for a p2p app ?

Kademlia was a homegrown routing algorithm, correct me if I am wrong. Some ideas may have been taken, but it's still the best routing algorithim for p2p.

[thongsai]

fakes is no big deal.. u just go into one of those verified links and ur fake problem is solved.. but corruption cannot be fixed.. except for another download.. but shouldnt it be right the first time? plus u guys think best coded means fastest dl and content.. 4 million users on any p2p network will make it fast.. if u guys dont like shareaza u should atleast try ares.. i mean ares users are as dedicated as shareaza users.. havent u guys wondered y we didnt join the kazaa bandwagon?

[Sephiroth]

Well alot of networks were left out and you cant see "the code" for some networks so there is really no way to tell.. A better way to have done it is what network is best designed and that is always argueable and it really depends on what you using it for and it changes with time so there is really no answer.

That and people here arent programmers.. Which really if you wanted to find what is "best coded" then you really have to compare the programs themselves. Which then it could be compared by how stable the program is or how efficient it is with CPU/Memory and how well it handles under stress and so on on how well coded it is.

Networks are not better than one another because they have fancy features or because they implemented one thing or not. In the end thats not what people care about.

As long as it works and is faster than others people dont really care how its done it could be with magical "file sharing faries" for all they care.

[John W. Lindh]

Which P2P network is best written in terms of speed, efficiency, not wasting bandwidth, scaling, etc?
I didn´t include Bittorent or DC because they´re too different.
And I´m really up for discussion here. G2 has been the most immpressive for me so far, but I´m not programer!

OpenNap (WinMX is OpenNap + their own network) is quite efficient but it's clearly not scalable since it is still relying on a centralized architecture. Any network using servers has a certain critical point where it stops accepting new users or completely crumbles under the load of traffic.

WinMX own protocol WP2P (that's its name IIRC) is not documented anywhere, I could only make assumptions how well it really performs, however since WinMX is not really developed anymore (despite Frontcode announcing a new version every now and then) I don't think it is one of the top networks.

Earthstation 5 is utter crap. It uses servers and will not be much different from OpenNap.

Edonkey2000 has large enough servers, so it can accomodate 2 million concurrent users, it's impossible to tell whether it will scale any further since it heavily depends on the servers. You might bring up eMule's Kademlia implementation at this point but that definitely will not scale infinitely and whether it will be able to replace the ed2k servers once they go down is questionable: client-server architectures are always a lot more efficient than p2p architectures and eMule's Kademlia implementation lacks some feature to balance the network load dynamically. If you are unlucky you may receive the queries for "porn" which may be more than you would like to handle (even if this traffic is distributed among 4 nodes AFAIK). You don't have any way of telling the rest of the network to go easy with their queries and the worst case would be that queries for "porn" or any other popular keyword don't return any results at all. ed2k's file transfers are quite inefficient, the last figures I saw were that the overhead for transferring files (without counting the traffic the servers had to handle) was 5-10% of the transferred content.

Fasttrack scales infinitely. No matter how many users, it will accept them. It is also very fast. Unfortunately it is not very efficient and you don't seem to be able to search more than 90-150k users. It also lacks some important features to speed up transfers like for example partial file-sharing and the ability to fix corruptions during the download without having to re-download the whole file.

Whether G2 will ever scale remains to be seen. If the clients really go for that global search, as the protocol suggests, it will never ever scale. G2's search architecture is not very efficient. Using UDP for searches is certainly very fast if done the right way (which is not the case since G2's searches are kind of sluggish) but using a connectionless architecture for searching will not be very efficient. On Fasttrack for example, all you have to know is one supernode and you can search the network. With G2 you will need ~2,000 hub addresses to search the network. Transmitting these 2,000 addresses alone accounts for a significant amount of overhead, other network that do not use UDP do not have. (I could mention Gnutella here but that's another discussion).
G2's HTTP downloads are not as efficient as they probably could be but they are a lot more efficient than ed2k for example.

You could have included DC in your list. DC does not scale but it is very efficient like OpenNap.

Bittorrent, like OpenNap, does not scale. A single tracker easily breaks down under the amount of requests it receives and hosting a tracker can eat up a lot of bandwidth. There have been some improvements and BT2 is supposed to offer some more fixes, but so far I have my doubts. BT is however quite efficient even if a lot of bandwidth is wasted on TCP overhead. It should perform about as well as G2's HTTP downloads.

Which network is the best, techology-wise, is a good question. Gnutella, which you did not mention, is certainly the most complex network of them all, and very powerful if implemented properly. BT is a very simple protocol to implement and quite efficient but it lacks some basic features like searches.Overnet, which you didn't mention either, is certainly also one of the best networks around. In my personal ranking Gnutella would be top but I'm clearly biased, being involved in the development of a Gnutella client. Overnet and Bittorrent would on second and third and I would put edonkey and G2 on 4th and 5th rank.

[stecbine]

Gnutella2 is not the best coded p2p stack at all. It has proven to not be scalable. It's dead, fasttrack has always been the best coded, then i would say that overnet follows. Fasttrack does need PFS but besides that it is very efective and proven scalable, G2 transfers just flat out suck. You can't beat kademlia nor fasttracks routing. G2 is based on the old gnutella routing alogorithm, perhaps minor changes and G2 downloads suck. WASTE is just IM with file transfers, it has no routing and is not really a network, it is a group of a couple of people talking directly. IMO
Oh man it's obvious you have a beef with Shareaza & or G2, because the garbage you just spewed made absolutely no sense, you backed it up with nothing and made yourself sound silly with your fanboy/hateboy comments. :tol