[LINK] docking the internet's tail

[Richard Chirgwin]

Kim Holburn wrote:
> It has been for some time (2004 is a long time ago on the internet) a  
> dream of the internets, that it would allow people to get access to  
> all sorts of exotic, esoteric, non-mainstream stuff:
> 
> http://www.wired.com/wired/archive/12.10/tail.html
>> The Long Tail
>> Forget squeezing millions from a few megahits at the top of the  
>> charts. The future of entertainment is in the millions of niche  
>> markets at the shallow end of the bitstream.
> 
> There was some debate last year started by Anita Elberse:
> http://hbr.harvardbusiness.org/2008/07/should-you-invest-in-the-long-tail/ar/1
> http://www.newscientist.com/article/mg20026873.300-online-shopping-and-the-harry-potter-effect.html
> 
> debate:
> http://springboardmedia.blogspot.com/2008/07/long-tail-not-so-debunked-after-all.html
> 
> Naturally it's difficult because getting statistics about something  
> that might by definition not have very much data is not necessarily  
> going to lead to accurate results.
> 
> Now here is an article showing that the "blockbuster" effect (the  
> opposite of the long tail) works in P2P networks.
> 
> http://www.wired.com/epicenter/2009/05/report-challenges-long-tail-theory-on-p2p-networks/
> 
> One thing everyone is happy using P2P for is finding out what's  
> popular.  But protocols like bittorrent tend to skew data away from  
> the long tail.  Bittorrent works better the more popular a thing is.   
> The more people interested in a piece of data, the more torrenters,  
> the faster the file moves.  Files that are rare and have few people  
> interested in them, may as well be on a single server, bittorrent  
> doesn't add anything to help, in fact it may even hinder compared to  
> say html or ftp.
> 
> Kim

Wow ... funny thing, I wrote something anti-long-tail fairly recently, and
copped a right shellacking from the various kinds of true believer.

But it never even remotely occurred to me that there might be a technical effect
at work! A sort of protocol feedback loop ...

RC
> 
>