home .. forth .. colorforth mail list archive ..

Re: Re: [colorforth] DARPA takes aim at IT sacred cows


From: kbk@xxxxxxxxx (Kurt B. Kaiser)
> Is anyone using 'bit fountains'?  

I recall one company selling software to do it, so I'd 
suspect someone's buying :-). But I don't know.

BitTorrent is somewhat reminicent, but it still uses 
packets. Designing a BitTorrent system that used ECC would
be (to say the least) interesting. Google to learn about
BitTorrent if you don't know what it is; I'm going to get
technical and assume that you know.

The current problem with BT is that getting the last part 
of a file is very slow, because many people turn off their 
clients as soon as the download completes, so there are few 
sources for the last parts of a file. With ECC you need to
have the entire file before you can start generating your
own unique ECCs, but you can always help out by parroting
the ECCs you've been sent. In the degenerate case (when
there's no users who have downloaded the actual file) this
means that only the server is generating ECCs, but as few as
one client has a direct connection to the server, and 
everyone else is connecting to other clients, thus sharing
bandwidth.

Hmm, seems practical to me-- and that's probably the worst
case. In reality, everyone who actually wants the file will
surely stay online until it finishes downloading, and then
when it finishes even if they're totally determined to 
contribute NOTHING they won't be able to hang up fast enough
to prevent their client from computing and sending a few
hundred new ECC packets.

>I imagine you would need a separate network to do this?  
>Or is there a transition solution?

Easy transition: just use the network you have, and send the 
ECC chunks as stateless packets. UDP/IP works well over the
Internet, for example.

>It's true there are interactions between the OSI 
>stack layers, but at least there is the intention 
>to minimize them.

The OSI model is nothing but philosophy; it's empty talk
and categorization, distinctions without differences.
There's no hard math behind it and no practical experience.

It makes a nice way to split things up to teach students,
but there are other ways, and there's no reason to suppose
it's anything close to the best, either for teaching or for
working.

> KBK

-Billy



---------------------------------------------------------------------
To unsubscribe, e-mail: colorforth-unsubscribe@xxxxxxxxxxxxxxxxxx
For additional commands, e-mail: colorforth-help@xxxxxxxxxxxxxxxxxx
Main web page - http://www.colorforth.com