Innocent until Client Analyzer proves them leecher

[Bebabob]

Client analyzer is one of the most important reasons I departed with the official client and came to emulefuture (initially through the beba mod and now using sharkx). By reading Wizard's postings on the forums, especially on the official forum, I get the feeling that Client Analyzer is Wizard's most precious brainchild, the project he is most proud of and justly so considering the achievement: heuristics against leechers. But I also believe there is room for improvement.

I am presenting a hardcopy of today's known clients screen, ordered by data transferred up. This is a typical screen which I see in all those 2-3 years I am using CA-enabled clients. It shows that Client Analyzer does identify leechers in general with a few misses. But considering that these clients have taken the most advantage of my upload, it also shows that perhaps I am uploading more than I should to leechers:



I know that I could avoid this by downloading from a large number of sources which could possibly turn me into a leecher, but I am getting this elsewhere. I am not having a problem with uploading, on the contrary, but I believe that uploaded should be given to people who haven't yet been proven leechers. For example to these folks:



The list above is again from the known clients screen, ordered by download status to get only people who are queued. None of them is a leecher; data are obviously inssuficient to mark them so. And in my opinion, giving these people a chunk is much more beneficial to the network than wasting upload on proven leechers. Yes, many of the new clients may prove to be leechers in the long run; but for the few who don't, that chunk may be their entrance to a beneficial tit for tat.

I don't know how this can be achieved. Heavy penalization of leechers ? Decrease of waiting time's importance on the queue priority ? Introduction of a new credit system tailored to Client Analyzer ? I would be really interested in everyone's opinion on the subject.

All data are taken from SharkX 1.7b3 with the following usage statistics:

Code: Alles auswählen

[b]Collected by bebabob using SharkX v1.7 BETA 3 (based on eMule0.49c)[/b]

[b]Basic System Infos:[/b]
[b]-------------------[/b]
|--> OS : Windows XP Professional Edition Service Pack 3
|--> CPU Type : Intel® Core™2 CPU         T5500  @ 1.66GHz
|--> Number of cores: 2
|--> RAM (Usage): 3.25 GB (69.28 MB)
|--> Free Space on Tempdrive: 45.96 GB

[b]Options:[/b]
[b]--------[/b]

[b]Emule Settings:[/b]
[b]---------------[/b]
|--> Max. Upload-Speed: 112.00 KB/s (Limit: Unlimited)
|--> Max. Download-Speed: 400.00 KB/s (Limit: Unlimited)
|--> HighresTimer: Disabled
|--> Max. connections: 800
|--> Max. new connections / 5 secs.: 80
|--> Max. half open connections: 36
|--> TCP Port: 80 UDP Port: 80
|--> Upload SpeedSense (USS): Enabled
|--> Connection Type ED2K / KAD : Disabled / Enabled
|--> Server Safe Connect: Enabled

[b]Mod Settings:[/b]
[b]-------------[/b]
|--> UPnP: Disabled
|--> Credit System : Use ClientAnalyzer embedded reward/punishing
|--> ConChecker: Enabled
|--> Retry failed TCP connection attempts No
|--> Save/Load Sources: Enabled (100|25)
|--> Max Upload slots: 30

[b]Upload Stats:[/b]
[b]-------------[/b]
|--> Upload Time: 18:11 Hours (100.0%)
|--> Upload-Speed: 88.55 KB/s
|--> Average Uploadrate: 91.96 KB/s
|--> Max Upload Rate: 100.46 KB/s
|--> Max Average Upload Rate: 0 B/s
|--> Uploaded Data: 5.74 GB
|--> Average Uploaded Per Session: 5.73 MB
|--> Active uploads: 23
|--> Waiting Uploads: 688

[b]Download Stats:[/b]
[b]---------------[/b]
|--> Download Time: 18:11 Hours (99.9%)
|--> Active Downloads (chunks): 8
|--> Download-Speed: 27.09 KB/s
|--> Average Downloadrate: 55.51 KB/s
|--> max. Downloadrate: 130.70 KB/s
|--> Max Downloadrate Average: 0 B/s
|--> Downloaded Data: 3.47 GB
|--> Average Downloaded Per Session: 7.82 MB
|--> UDP File Reasks: 7.58 k, Failed: 962 (12.7%)
|--> TCP-connections: 4.10 k, Failed: 0 (0.0%)
|--> Found Sources: 275
|--> Too many connections: 0
|--> On Queue: 224
|--> Asking: 0
|--> Connecting: 1
|--> No needed parts: 22
|--> Asked for another file: 5
|--> Queue Full: 20
|--> Cannot connect LowID to LowID: 0
|--> Banned: 9
|--> Filtered: 455
|--> Problematic: 0 (0.0%)

|--> eD2K: 932 (92.1%)
|--> Kad: 947 (93.6%)
|--> eD2K&Kad: 868 (85.8%)
|--> Unknown: 1 (0.1%)

[b]Statistics:[/b]
[b]-----------[/b]
|--> Session UL:DL Ratio: 1.66 : 1
|--> Total size of Shared Files: 32.93 GB
|--> Number of Shared Files: 52
|--> Number of Downloads: 3
|--> Completed Downloads: 0
|--> Secure Ident (OK : Failed ): 1002 (99.0%) : 5 (0.5%) : 5 (0.5%)
|--> Download Sessions: 463
|--> Successful Download Sessions: 454 (98.06%)
|--> Failed Download Sessions: 9 (1.94%)
|--> Upload Sessions: 1260
|--> Total successful upload sessions: 1026 (81.43%)
|--> Total failed upload sessions: 234 (18.57%)

|--> *** Modded by taz ***

[Tuxman]

Heavy penalization of leechers ?

Putting them on the bottom part of the queue is a heavy penalization.

[Bebabob]

Putting them on the bottom part of the queue is a heavy penalization.

Then why it is the leeches that get the lion's share of my upload ?

I am taking another screenshot to make this clear; this is the top of my known clients list:



So most of my uploaded data goes to leechers. Practically everyone of those highly benefited clients is on my queue and he is a leecher!

Obviously getting them on the bottom part of the queue is not effective enough...

[Tuxman]

The known clients list is not of any interest here. The On Queue list is.

[myth]

And look! The "leechers" gave you next to how much you gave them!

[Bebabob]

The known clients list is not of any interest here. The On Queue list is.

See the "download status" column. All leechers are queued.

Please take the time to read what I've written, I think you are overlooking some of the information.

[Tuxman]

Again: Queued vs. "somewhere in the bottom part of the queue".

[Bebabob]

And look! The "leechers" gave you next to how much you gave them!

The vast majority of them have not. They are real leechers. The rest are thus due to Wizard's heuristics which I find quite acceptable.

But the point is that the vast majority are typical leechers, by definition, who got more than what they have given.

[Bebabob]

Again: Queued vs. "somewhere in the bottom part of the queue".

How does bottom part of the queue matter if they are getting what they want ?

[Tuxman]

As long as they don't enter your upload slots prior than any "better" client with more score, everything is fine with the CA.