🚀 go-pugleaf

In <3lg0b5$ae@lace.Colorado.EDU>, crosby@nordsieck.cs.colorado.edu (Matthew Crosby) writes:
>In article <>, sle@sensemedia.net <sle@sensemedia.net> wrote:
>>Check out this url: news:jnavasD5p92C.G0A@netcom.com

>>graphics, web and video do not even raise a candle to the amount of spam on
>>the net purely from the innefficient routing of news via usenets archaic protocol...

>>... feeds filling up the 'superhypeway' with redundant data to be stored on
>>countless disks, and each site not actually utilizing more than a few percent
>>of the total feed.

>Not even close to being true.  Usenets store and foreward topology scales
>MUCH MUCH better then a store in one place topology like the Web or FTP.

True.  Just consider how difficult news reading would be if all usenet users
were all trying to get into one server somewhere.  It is bad enough in a "small"
special interest area like OS/2 Internet Software when a new version of a much
anticipated software release gets posted on hobbes.nmsu.edu.

In fact, for heavily used services like news, the more local the news server, the
better the service to the customer....  It would be interesting if WWW went that
way, but the size of the graphics files is the killer.

Still, if somebody is looking for an interesting research project, how about a
WWW cache memory server that stores frequently accessed WWW pages....

>Consider:  If I have a site, and I have a feed for something, it only
>takes two readers and I already have a 50% reduction in used net bandwidth
>if I have a local copy then if I read it remotely.  It takes up disk,
>but disk is very cheap--cheaper by far then network bandwidth.

Ummmm, not really.  A gigabyte of net backbone bandwidth costs about $0.75
while a gigabyte of hard disk costs about $400.00.  So, hard disk space is 533
times more expensive than the equivalent amount of net bandwidth.

Of course, this is the tyical totally irrational result one gets from comparing
apples to oranges.  (For the technically interested, I used the Pac Bell ATM
tariff for net bandwidth, which is 155Mb/s for $32,000/mo in order to arrive
at the roughly $0.75 per gigabyte figure.)

Did you spot the fallacies?  A gigabyte of net bandwidth at 155Mb/s lasts only
about 52 seconds, while a hard disk will last for several years.  Also, the
gigabyte of net bandwidth is a "use once" resource, while the hard disk can
be used over and over again, based upon the expiration algorithm of the
contents of the hard disk.  I am sure there must be others.....

>Store and foreward also has other advantages, for example it doesn't have
>to be real time.  And how about all them folks still running UUCP?
>
>It has some disadvantages, of course, the most pronounced being propogation
>time.

Propagation time is one of the issues which somebody could "take charge of" 
and improve.  I am certain that a decent N-Space modeling algorithm could
lop an order of magnatude off of the typical propagation delay.  The real
issue is that nobody wants to "take charge" of fixing something that is only
"broken" from a purely cosmetic standpoint.

>I suspect the best solution is a compromise,ie something like a single site,
>but  with lots of cacheing.  This sort of thing is already being loked at 
>(eg: the Harvest project)

It should be forgotton.  The rate of growth for Internet accounts should explode
over the next year as every major on-line service and telephone company jumps
into the Internet business with as much force as each can muster.  If such an
idea is only marginally practical today, you can forget it for tomorrow.

As I mentioned above, you just need to do a decent N-Space model of the
news-neighbor situation, allowing for the 45Mb/s ATM backbone to get high
speed propagation between areas of the US of A, and tell the rest of the
world to upgrade their backbone(s) to keep up.

>(oh, and compare the amount of bandwidth used by text versus that used
>by video some time!)  

YES!  A decent compressed video channel for a standard cable television feed
(such as appears on the 18" dishes) runs about 6Mb/s!  That drowns most
Internet stuff.  So, how would the Internet support video on demand?  The
same way it supports news now.  You have a local video on demand server
that gets loaded once over the net with the video signal, using TCPIP because
it does not have to be loaded in real-time (i.e., basically an FTP situation).  The
users access the video off of their local servers, just as they do news.  Its the
only practical solution at the present time.

---------------------------------------------------------------------------------
BILL     |   But my mother doesn't let me actually have these opinions.