Re: [freehaven-dev] Opening example

[Roger Dingledine <arma@mit.edu>]

On Fri, Nov 17, 2000 at 11:11:25PM -0500, dmolnar@belegost.mit.edu wrote:
> I was thinking of the Napster bandwidth usage problem. Harvard recently
> started filtering incoming Napster connections to reduce network load; the
> local admins claim that at one point a large proportion (20%) of our
> bandwidth was spent on Napster transfers to other sites (since it sets up
> a machine to work as a server automatically). Other universities have
> cited bandwidth usage as a reason for banning Napster.
> (The question is whether this is a fig leaf for avoiding the RIAA...)

I think this is better than the slashdot example, at least. I'm not sure
if it's the best example we could use. I'm still pondering that one. I
think Napster is pretty good: it's the poster-child of the book
and everybody reading the book will want to hear more about napster.
 
> Another problem - Gnutella freeloading. The recent study claiming that
> a small percentage of the people provide most of the MP3s. There's one
> person who thinks the study is flawed, however- his response is at 
> http://www.infoanarchy.org/?op=displaystory&sid=2000/8/22/33519/1404
> for what it's worth. 

I didn't read the original study in detail. But there is no question that
Gnutella has been having serious growing pains, resulting in making it
suck to actually get files out of it these days.
 	
> Does anyone actually use Gnutella here? What are usage anecdotes like?

I used it for a while. Used to be that I could do searches and get some
responses that weren't bogus. As of a month or two ago, I generally
don't get any responses from my queries. I think this is because my
original "friends" list which I downloaded long ago is no longer very
useful. I have no idea where to get a new friends list. My client is old;
I assume newer clients have better ways of picking friends for me. But
this illustrates the overall problem that both Freenet and Gnutella have
(and haven't sufficiently solved yet) -- introducers are very difficult
to get right.

Speaking of, we really haven't spec'ed out a detailed proposal for
introducers for Free Haven yet. We criticize Publius for not having any
real support for a dynamic system, whereas we don't have much of a system
at all yet. :)
 
> What other major problems could we open a chapter with?

We could emphasize Gnutella more. But I don't think we want to harry that
poor Gnutella guy any more than we already have, and I think Napster and
Gnutella are in a similar situation and Napster is much higher-profile.

One of my arguments against using Napster as the lead-in example is that
it doesn't address the idea of accountability for data flooding, which
is a major part of the chapter. (Or is it -- is it a major issue for
accountability but doesn't get as much air-time in the chapter?)

In order to describe that, tho, we need to use an example where files
actually get moved to other computers. Such systems are not widespread
in the public eye, currently. (True? Akamai doesn't count?)
(Look at http://www.oreillynet.com/pub/q/p2p_category for a sample
(incomplete) list.)

> -David
 
--Roger