[p2p-hackers] Re: Identify "defective" nodes

[Karl A. Magdsick]

You want to minimize the amount that you hurt network performance 
through ignorance
of protocol nuances and application internals.

It would be much better to be able to ask each application for sets of 
byte ranges that the
application has downloaded/is downloading and also be able to offer each 
application byte
ranges that it is allowed to download next.  You would also want a way 
to ask each
application for which byte ranges of the download have been verified by 
the application,
as well as a way to notify each application of the set of bytes that 
have been verified by
any application.

With two clients, presumably you would tell one that it is responsible 
for the first half
of the download, and tell the other application that it is allowed to 
fetch the second half of the
download.  As the download progresses, you would modify the boundary 
between the two
programs, making sure not to allow any program to download bytes that 
have been downloaded
by the other.

You'd still have to be very careful, or you could end up really hurting 
network performance.

You'd probably want a preview-accommodating client (such as LimeWire) to 
fetch the first half
of the download, and have a preview-indifferent client (such as most/all 
BitTorrent clients) fetch
the second half of the download so that the user could still preview the 
download.


-Karl

Daniel Stutzbach wrote:

>On Tue, Feb 28, 2006 at 01:56:43PM -0500, Greg Bildson wrote:
>  
>
>>>Yes, true, it would also need to be able to tell the application "go
>>>fetch this chunk" rather than letting the application choose.
>>>      
>>>
>>Which would likely undermine any algorithm trying to improve availability by
>>downloading random or most rare chunks.
>>    
>>
>
>Well, the plug-in could take over doing that part, too.
>
>I did mention this was my pipe-dream plug-in, right? :-)
>
>  
>