Re: Bandwidth

Daniel S. Riley (
25 Nov 1997 13:14:44 -0500

To: James Green <>
From: (Daniel S. Riley)
Date: 25 Nov 1997 13:14:44 -0500
In-Reply-To: James Green's message of "Tue, 25 Nov 1997 13:12:52 +0000 (GMT)"
Message-Id: <>
Subject: Re: Bandwidth

James Green <> writes:
> If I had wanted a system like caching, I wouldn't have suggested a 
> theoretical solution to the problem. I know what a cache does and how 
> to use one, but the high-traffic problem still occurs daily, so a 
> different solution was one that I was talking about. Forget caches 
> please!!!

But what you are proposing is a cache--the routers (or whatever) have
to store the pages they are going to serve, and that makes them
caches.  If we forget caches, we have to forget your proposal too.

> Whilst they do help enormously, they do not solve the problem entirely. 
> For instance, in a live broadcast, people can hardly get the entire 
> file from a cache went the EOF hasn't been transmitted yet, can they? 

There is no reason a cache can't start transmitting the beginning of
a document before it has received the end.  (And if there were a
reason, it would apply to your proposal as well, since your proposal
relies on caches.)

> A file (not cached yet as it is a new one) is now available on an 
> American site. As soon as this becomes apparent, people from Britain, 
> Netherlands, Sweden, France, Germany, Italy .... Australia, Japan, 
> etc., want it (it's popular). Seeing as half of these (I don't know 
> real statistics) aren't using a cache, the file is ordered by their 
> ISPs directly from America.
> Sure, a lot of Internet infrastructre would have to be re-configured, 
> and individual countries would have to have their ISPs co-operate to 
> provide centres rather than ISP shared lines going in to the country, 
> but it makes a lot more sense than the current 'just get the file' 
> system.

So basically, because people don't use caches now, you're arguing that
we should completely overhaul the worldwide IP infrastructure to force
them to do so?

> Now do you get what I am talking about? Maybe someone at Cisco (however 
> you spell it) is reading???

See <URL:> for information on the work on
national and international cache hierarchies.

See <URL:> for
what Cisco has been doing.
Dan Riley                               
Wilson Lab, Cornell University       HEPNET/SPAN: lns598::dsr (44630::dsr) "Distance means nothing/To me" -Kate Bush