|
FineCS try to force the user to payback as well (3 chunk gap threshold) but it dow not enforce it in any way. I belive that the thing has to find an equilibrium by itself.
If you download a lot, but you do not upload enough, your download speed will decrease. If you upload as a mad, or focus your upload to a single file (PS, part file push, friendly slot, whatever) which is the same, you still do not get anything immediately, but your upload "credit" will be diluted over time when you will download and will not upload. That's the idea of not using boost mechanism, but only punishing those that do not upload enough.
This way, you do not create strong TFT like mechanism but you are still rewarded if you upload. By using your system there are 2 easily exploits interdependent:
1. use all the bandwidth for payback. Once you admit the thing is "legal" the ratio of free bandwidth vs payback one is just a matter of opinion.
2. you'll be completely open to credid shaping. Expecially if you use the "upload but not request" idea.
In the end, if you use a 1/5 bandwidth payback you can only obtain, on average, a 1/5 more benefit of download... see point 1.
On the other end, you would "steal" 1/5 more bandwidth from rare files.
Now, about this mechanism, you can't think that the people sharing rare fies will not be put in disavantage or will continue as they are doing now.
If such a system would become wide spread, to obtain a download in a resonable time (or ever) you have to use the TFT mechanism (or you'll find yourself contantly pushed down in the upload queue).
Sharing those rare would not be of any particular advantage (as you would not expect the other one to have something for you.. it's rare!) and would "steal power" to previous process.
With this system upload get a value if you have popular incomplete file to share, as you would download them faster. Sharing rare, or even complete, would mean "loosing" value as that upload may only be of some use, while the payback one would have surely its value.
Hording, at whatever level, is not good. See what happened "at the other side". It's an example on how the tradeoff between file availbility and richness can be easily disrupted.
And I say it again, don't think that the 1/5th bandwidth limit will be ever respected one you'll see that breaking that would make your download really fast (is this the implementatio of zz's "download really really fast" 
). Now, 0-uploading obtain less that they could if they would use their upload sensibly. Credit shapers get just a small amount more (due to the way the official CS work, ufortunately). Your system would simply open the door to the latter, which may bring more upload to the network as a whole, but would, for sure, make it poorer than it is now.
I'm sorry. It's a long post, full of errors and with casual thoughts thrown here and there. But I'm tired and have to code a bit more before going to sleep 
This post has been edited by CiccioBastardo: 19 December 2005 - 08:24 PM
(责任编辑：)
|
