[ NNSquad ] Re: Is the bandwidth hog a myth?

[Wes Felter <wesley@felter.org>]

> Benoit Felten of the Yankee Group challenges the "myth of the bandwidth  
> hog" used by ISPs to fight network neutrality:
>
> http://www.fiberevolution.com/2009/12/whats-a-bandwidth-hog-.html
>
> He argues that TCP is inherently fair (though he does not address UDP).

This is definitely not true; TCP implements a pretty weak form of 
"fairness" that is different (and worse) than the common sense 
definition of the term. George Ou and Lawrence Roberts have explained 
this several times.

Even assuming fair bandwidth sharing, it's questionable whether it's a 
good idea for an ISP to run a link at 100% utilization which tends to 
increase latency.

I agree that we should be worrying about actual congestion rather than 
just assuming that the customers who transfer the most data are likely 
to be causing any congestion that may exist. (To throw in an anecdote, I 
have noticed virtually no congestion on my broadband for years; perhaps 
none of my neighbors are "hogging" and thus no one in my neighborhood 
should be punished.) This type of monitoring will likely be more 
expensive than simply counting bytes, so we should be careful not to add 
too much cost here (I wouldn't want the Internet to becoming like the 
PSTN, which is a giant accounting/billing system that also completes 
phone calls).

Interestingly, the much-reviled Comcast is actually a leader in this 
area, with their protocol-neutral bandwidth management that tracks 
actual congestion (or "near-congestion"). Perhaps once this system has 
been improved with experience (I hesitate to say "perfected") Comcast 
will find that caps are unnecessary and do away with them.

Wes Felter