[petermrg]

I know that multiple server connections are against emule policy but I strongly believe this should be reconsidered.

This feature, for example, would allow users to offer himself as source to a server given in the ed2k link (useful for website admins).

A sample elink:
ed2k://|file|filename.ext|12345678|A1B2C3D4A1B2C3D4A1B2C3D4|server,123.45.67.89|/

The user connects like always to his preferred server AND to the server pointed in the elink, so he is source in both servers. The normal client-server activity is done on his preferred server.

Desired features:

• The available sources for a file would be centralized on a server AND also available on the other servers (because the user remains connected to his favorite server).
  
• Won't cause much overhead because the additional server connection is only used to share the file
  
• Website+server admins can know the real sources for a file
  
• Works with LowID clients
  
• This is similar to torrent trackers
  
• A user can decide on which server(s) he is going to share his files (added security/privacy).
  
• In order to workaround max shared files limitation, the client should open connections to as many servers as required to share all files he has. Today, with terabyte disks it's easy to pass the limit.
  
• For security reasons, the user should have the option to manually set the limit of shared files per server.

[tHeWiZaRdOfDoS]

Multiple server connections and your proposed feature are not necessary because eMule can ask all servers for sources via UDP plus it can get sources via kad, xs and passively.
IMHO, this wouldn't have any benefits.

[DavidXanatos]

Server UDP as well as Kad is limited to file's with less than 50 sources only.

Also server callbacks are cheaper than Kad Callbacks so a LowID sitting on more than one server would be more useful to the network.

So such a feature would be quite nice to have.

David X.

[tHeWiZaRdOfDoS]

There's no problem finding sources for popular files... IMHO, that'd be wasted OH.

[DavidXanatos]

Well, if you want decent download speeds, you need with eMule even for popular files as much sources as you an get.

For the modern every day user it is important the download starts right away, and this requiters to get as much sources as ASAP.

When in doubt look how BitTorrent is doing it and do it.

[tHeWiZaRdOfDoS]

DavidXanatos, on 08 April 2013 - 10:01 AM, said:

When in doubt look how BitTorrent is doing it and do it.

That is a pretty stupid advice, even for you.
This. Is. eD2K. Not. Bittorrent.
You should learn to live with that 'cos it's a fact.

eMule does NOT waste overhead (yet) and it's NOT a TFT client.

About the sources: usually you will get some sources to start with either from servers/kad or both, ask via XS and you will have even more sources.
Have a little patience and the other clients will find you and you will add them via passive src finding.
It's not the quantity of sources that matters but the quality, the problem today is that most ppl either use leecher mods straight away (... for "security purpose", of course), exploit the stupid built-in ratio of eMule (i.e. giving 10kB/s when they could and should give way more) or have simply misconfigured their client.
Anyways, you usually end up with a few hundred or even thousand clients who don't give a single byte while sucking you dry whilst there is usually only a handful that actually helps you completing your files.

So IMHO the way to go is NOT to find more sources but to improve the quality of the existing sources, i.e.

• fight leechers
  
• change the ratio system
  
• make emule easier to configure

[DavidXanatos]

tHeWiZaRdOfDoS, on 11 April 2013 - 06:53 AM, said:

This. Is. eD2K. Not. Bittorrent.
You should learn to live with that 'cos it's a fact.

eMule does NOT waste overhead (yet) and it's NOT a TFT client.

And eMule suxxx a lot in comparation to BT.

You should think about why Bittorrent is so popular, successful, and provide such a satisfying download experience, and eMule don't.

The main Reason is they don't give a damn about overhead and instead of fighting leecher they work on making downloads complete as soon as possible. (including even wistfully redundant requests for the last few blocks, to avoid slowing down file completion on the very finish)

Sure they "waist" bandwidth, but they also distribute it in a way that it does not negatively affect in a _noticeable_ manner the download speed.

Quote

Have a little patience and the other clients will find you and you will add them via passive src finding.

The modern user does not have patience,
how do you justify to user the need for patience when other systems like BT don't require any patience at all, usually you can start watching your downloaded movie on the fly as it trickles down on your HDD.

Quote

leecher mods straight away (... for "security purpose", of course)

This argument applies equally to Bittorent, same legal situation same incentive to use 0-upload mods in many EU countries where the pure DL is 100% legal.

Quote

exploit the stupid built-in ratio of eMule (i.e. giving 10kB/s when they could and should give way more)

Imagine that, transmission (the main bt client for linux) does not have a ratio at all.

Quote

So IMHO the way to go is NOT to find more sources but to improve the quality of the existing sources, i.e.

Well BT is very aggressive when find more sources, they basically flood their PEX (Bt for XS) messages _every minute_ and include all new sources found since the last PEX message sent to peer. So their aim is to make every peer on a particular file know every other peer, asap.
No waiting for passive sources or shit...

Ultimately for any reasonable user a client that starts upload right away but makes tham wait and wait and wait for the download to complete is unfair.
You may don't like this attitude and you may despise this users, but as thy make up the very majority you must accept it and adapt.

eMule should drop the queue nonsense, and start uploads randomly as Bit Torrent does.
This way your download speed will not depend on your up time, and thats very important for a good download experience.


I mean yea, THIS IS SPAARRRTTTTAAAAAAA.... eee.. no eD2K!!!!!!
as you said or so...

But still doesn't the staggering performance and popularity of BitTorrent make you think even for a moment, that you may be doing something wrong with you current (outdated) policy?

I mean no Mather how perfect something sounds in theory (like communism for example) if if fails miserably in the real life it has to be rethought and possibly some core ideas even thrown over board.
What good does a client gives you that has the smallest overhead and most restrictive leecher fighting schemes and the coolest grammar in its protocol, or whatever ...
When it is used only by a hand full users, And surpassed in its popularity by something so trivial like BitTorrent?

Look at the development of the last years:
2006: BT 53%, ed2k 49%: http://www.ipoque.co...g-boom-p2p-file
2007: BT 66% ed2k 28%: http://liquidculture...p-file-sharing/
2009: BT: 70% ed2k 24%: (for germany): http://torrentfreak....traffic-090218/
2011: BT 74% ed2k 7,5% http://broadbandbrea...study-suggests/
Now the BIG question will ed2k have an own bar in 2013? or will it be swift under the rug as other with the obscure rest?


Don't get me wrong I like eMule (may be more due to sentiment than rational reasons), but still I like it and I really wouldn't want it to become a part of "others" (together with such successfull networks liek ants, or monolith or retro share, or whatever).

I want eMule to be used by as many users as possible and in order to achieve this eMule must once again become appealing to the masses.

And being appealing to the masses means the download starts when the user want it to start, and the user does not have to "wait" for a popular download to complete.

Without BT being around you would probably clame that thats impossible with P2P, but we have BT and it proves that it is very well possible.

So it does not Mather how much the needed changes go agains your personal conviction, they are needed and proven to be successful in real live (BT).


Overhead was important 10 years ago, when you head only 128 kbit/s upstream at all.
Today the majority of users have > 1MBit/s overhead becomes a meaningless design aspect, this is a fundamental paradigm shift for P2P development indeed. But one that must be embraced and used, instead of clinging to 10 year old design policy's that simply are not longer valid in the modern world.


Lets look on your second favorite design paradigm, fairness and leecher fighting.
BT does not give a damn about leechers their policy is make the client apealing enough to the masses so that those who give are enough to support all. It is a very socialistic approach but yet once again one that proves to be Superior.
Instead of trying to force every user to give as much as he can, they allow every user to give as much or little as he wants hoping that there will be enough users to give enough for the whole network, its a bet on human nature and effects of large masses, but obviously one that BT's designers won hands down.

Its quite simple actually, if you are in a environment where to support all, all need to "work" you can not afford to allow people to be "lazy" still get support, in a environment however where whats needed is abundantly available you just have to get some people to "work" a little bit to get support for all.
Its kind of like the German BGE (basic income guaranty) discussion, if we know we have the resources to support all while allowing only those to work who want voluntarily, there wouldn't be a discussion at all. The discussion basically is not should be, but can we already. And in real life politics thats not so easy to answer.
But in our P2P case, BT just went ahead tried it and yes we can! Our pipes are brought enough to have voluntarily seeders provide enough download for all leechers.
Is that fair? Sure it is, the "exploited once" are voluntaries, they want it this way.



David X.

[petermrg]

Totally agree with David.

tHeWiZaRdOfDoS said:

This. Is. eD2K. Not. Bittorrent.

Right. But why not learn about the good parts of BT and apply them to ed2k? (uTP, for example, is a MUST).

DavidXanatos said:

Look at the development of the last years:
2006: BT 53%, ed2k 49%: http://www.ipoque.co...g-boom-p2p-file
2007: BT 66% ed2k 28%: http://liquidculture...p-file-sharing/
2009: BT: 70% ed2k 24%: (for germany): http://torrentfreak....traffic-090218/
2011: BT 74% ed2k 7,5% http://broadbandbrea...study-suggests/
Now the BIG question will ed2k have an own bar in 2013? or will it be swift under the rug as other with the obscure rest?

This is very sad, but it's normal because new developers or new ideas aren't welcome.

[tHeWiZaRdOfDoS]

DavidXanatos, on 11 April 2013 - 09:30 AM, said:

You should think about why Bittorrent is so popular, successful, and provide such a satisfying download experience, and eMule don't.

The reason is simply: because BT exchanged file diversity for speed. Also, and that's *really* a problem with eMule: it's FAR easier to use and understand for n00bs... and from my experience about 90% (and up) of all ppl accessing the internet ARE n00bs. Today, every idiot can use the internet on thousand devices because it became so easy... unfortunately, they do
Anyways, that's why I wanted to release a "n00b-friendly" client with kMule... even though it's still far from perfect, it's way easier to use.

Quote

The main Reason is they don't give a damn about overhead and instead of fighting leecher they work on making downloads complete as soon as possible. (including even wistfully redundant requests for the last few blocks, to avoid slowing down file completion on the very finish)

Ahem... that's just plain bullshit. True, they hammer their requests and they don't fight leechers per se but as you know very well, you need trackers for most downloads and they do fight leechers. Of course they do. BT is successful because it concentrates on a FEW files at a time while showing its users that they actually get something for their upload.

Quote

Sure they "waist" bandwidth, but they also distribute it in a way that it does not negatively affect in a _noticeable_ manner the download speed.

Well, you know quite well that BT (I don't mean your self-built stuff) doesn't feature a built-in search/indexing system which would rely on other clients being online and being careful with the OH considering that a huge network relies on it.

Quote

The modern user does not have patience,
how do you justify to user the need for patience when other systems like BT don't require any patience at all, usually you can start watching your downloaded movie on the fly as it trickles down on your HDD.

Granted. Anyways, for a popular file, I usually get a few 1000 sources within less than a minute AND it starts downloading (fast!), too...
The chunk selection in eMule isn't suitable for streaming, true, but it ensures fast and reliable file spread even after some days.

Quote

This argument applies equally to Bittorent, same legal situation same incentive to use 0-upload mods in many EU countries where the pure DL is 100% legal.

Right, but trackers will keep them away... alright, I know, there are some "free" trackers but they usually feature only some "free" files.

Quote

Imagine that, transmission (the main bt client for linux) does not have a ratio at all.

Yeah, but, trackers... blabla - always the same as above.

Quote

eMule should drop the queue nonsense, and start uploads randomly as Bit Torrent does.
This way your download speed will not depend on your up time, and thats very important for a good download experience.

I agree that this would really be an interesting experiment which may prove successful, at least in the long run. I remember SF doing some maths about that a few years ago.

Quote

I mean no Mather how perfect something sounds in theory (like communism for example) if if fails miserably in the real life it has to be rethought and possibly some core ideas even thrown over board.
What good does a client gives you that has the smallest overhead and most restrictive leecher fighting schemes and the coolest grammar in its protocol, or whatever ...

True, but remember that eMule and BT have 2 totally different goals. Given, it may be - due to the changed used attitude - that eMule isn't use that much anymore but it's no use to turn eMule into a BT clone... BT exists already.

Quote

BT does not give a damn about leechers their policy is make the client apealing enough to the masses so that those who give are enough to support all. It is a very socialistic approach but yet once again one that proves to be Superior.
Instead of trying to force every user to give as much as he can, they allow every user to give as much or little as he wants hoping that there will be enough users to give enough for the whole network, its a bet on human nature and effects of large masses, but obviously one that BT's designers won hands down.

BT works via TFT, what you describe would be eMule's philosophy...

[Zangune]

It would be interesting to estimate how fast is source finding and how much bandwidth is ''wasted'' because of leecher mods and 10KB/s exploiters.
I guess leech mods percentage is low, am I wrong?

[tHeWiZaRdOfDoS]

From my experience it depends on both the files you want to download (usually most leechers download common/new stuff) and how exactly you're defining a "leecher".
I think - that's my opinion and a wild guess - that about 90% of all eMule users don't use eMule with the maximum possible UL speed.
Granted that there are quite a lot out there that are limited by their ISPs and/or don't have flat rates... though ppl seem to be willing to pay for certain downloads/hosters/trackers/whatever but don't care for a better line enough to annoy the hell out of their crappy ISPs.

About leechers... it depends. I downloaded files with no leechers at all (though a lot of GPL breakers) and then again I had almost 100% leechers.
Another problem is that eMule isn't designed for seeding. Disk and CPU usage usually go through the roof and you can't use the full upload capacity of eMule when run on some server - though AFAIR SomeSupport mentioned that the next version of eMule will have improvements in that regard.
Anyways, that fact and the platform problem (though eMule basically runs fine on wine it's not designed to do that) prevent eMule from being used by seeders like it's done with BT clients. Most servers run Linux of some kind and a native Linux application (yeah, I know, aMule - ever tried to use it?) would be superior.

[DavidXanatos]

tHeWiZaRdOfDoS, on 11 April 2013 - 06:48 PM, said:

The reason is simply: because BT exchanged file diversity for speed.

In my experience you can find currently more stuff on tpb or through google (OCH) than in emules search.
Sure it wasn't always like that, but no users means no files.
So if you loose users due to bad speed you are exchanging file diversity for nothing.

tHeWiZaRdOfDoS, on 11 April 2013 - 06:48 PM, said:

Ahem... that's just plain bullshit. True, they hammer their requests and they don't fight leechers per se but as you know very well, you need trackers for most downloads and they do fight leechers.

Bullshit, you can download nowadays with DHT instead of trackers just as good.
And with somewhat popular stuff you will have enough seeders to be able to download only from then. So no need for a lot of Upload and piece trading with other downlaoders.
With any peer in the swarm that already completed the file you can not do any traiding anyways and those get decent speeds.

tHeWiZaRdOfDoS, on 11 April 2013 - 06:48 PM, said:

Right, but trackers will keep them away... alright, I know, there are some "free" trackers but they usually feature only some "free" files.

Thats bullshit again, almost all popular torrents nowadays use tracker.openbittorrent.com or tracker.ccc.de booth are completely open. And as mentioned above you can use DHT.

Quote

Yeah, but, trackers... blabla - always the same as above.

trackers ate not magic bullets against leeches - always the same as above.

Sure there are private Anti Leech Trackers out there but they don't share public torrents, but torrents created with a private flag set, you can not find such torrents on TPB or elsewhere in the public web. This are private communities and for the global file sharing completely irrelevant.

Quote

BT works via TFT, what you describe would be eMule's philosophy...

That is so wrong! As explained earlier BT uses TFT only between clients that booth have an incomplete file.
Seeders upload to anyone completely egalitarian, not like emule with its FiFo queue that benefit users that run their client 24/7 and punishes casual down loaders.

David X.

[Sir_Boagalott]

petermrg, on 11 April 2013 - 05:30 AM, said:

Totally agree with David.

IMO they are both right on several issues. However it appears that Xantos has used bt more than WiZ.

All bt sites have trackers, but not all trackers care aboot ratios. The majority of decent private sites have ratios, or at least track shared time so you cant hit n run. i.e. unshare completed files right away. With most public sites, sure they track ratio but it means nothing, you can definitely leech away and h-n-r at will.

Personally I like the TFT idea for ed2k on incomplete files only. I've suggested that many times, same with another similar lil simple trick bt does: a 100MB line UL to you a bunch, so you owe them a bunch, you UL a lot to them back, which actually makes use of their 100MB line and spreads files quickly to everybody. eMule treats clients way to "equally" even when they are often clearly not equal. Personally I have never been a big fan of eMules score system.


If Xantos knows bt so well, and wants ed2k to work more like bt, what is 1 of the primary reason as to what makes bt faster? So far he has eluded to say. Ironically WiZ nailed it.

In theory it should be possible to intelligently do both - speed with file diversity.