Idea #3: A peer to peer backed video website

[ru_linux_geek]

This idea competes with Idea #2: Transcode text in videos much better. If YouTube won't deliver the optimal StarCraft video over the internet watching experience, why not design something that can? Actually, this idea is a lot more like Joost than YouTube, except minus the heavy handed, traditional dichotomy between content consumers and content

Comments

Re: You don't even have a screen shot.

[anonymous]

I think it would be fairly easy (or at least straightforward) to write a program that extracts single frames (equivilant to screenshots) from a sparse regular file, like what BT generates.

I really don't understand how your concept above significantly differs from how BT works in practise. Also, I think that the only way to reward prolific sharers on a non-per-file basis is for each client to trust to a third-party, but that makes abuse easy.

-Dave Harding

Re: You don't even have a screen shot.

[ru_linux_geek]

What is wrong with the master keeping track of reputation, and all the clients trusting the master? This is already the model implicit with the bittorrent tracker, isn't it?

Re: You don't even have a screen shot.

[anonymous]

No, the tracker doesn't keep track of reputation authoritatively now. How excactly is the BT tracker supposed to keep track of reputation in the current model[1]? All the data is exchanged between peers. If the tracker trusted a peer to report its reputation, the peer could lie. Even if the tracker only trusted multiple reporting peers (the uploader and the downloader), the peers could be in lying colusion -- or they could simply have some unfair advantage, like being on the same local network.

[1] Or even your model?

Re: You don't even have a screen shot.

[ru_linux_geek]

Maybe it's harder than I originally thought :).

What if you operate under the assumption that most peers will be honest and that initially, all the peers will be honest? Then you can gain trust as the system operates, and so long as there isn't collusion by very trusted peers, it seems that believing the most trusted peers will work fine.