Path: utzoo!yunexus!geac!syntron!jtsv16!uunet!lll-winken!lll-lcc!ames!ncar!mailrus!tut.cis.ohio-state.edu!uccba!uceng!dmocsny
From: dmocsny@uceng.UC.EDU (daniel mocsny)
Newsgroups: comp.sys.next
Subject: Re: NeXT's BIG 3.5" mistake.
Summary: Storage reqs. vs. access times; information reqs. of knowledge workers
Message-ID: <355@uceng.UC.EDU>
Date: 29 Oct 88 16:11:41 GMT
Article-I.D.: uceng.355
References: <0XMtqn087E-0A14EYk@andrew.cmu.edu> <344@uceng.UC.EDU> <5772@hoptoad.uucp>
Organization: Univ. of Cincinnati, College of Engg.
Lines: 73

In article <5772@hoptoad.uucp>, tim@hoptoad.uucp (Tim Maroney) writes:
> In article <344@uceng.UC.EDU> dmocsny@uceng.UC.EDU (daniel mocsny) writes:
> > A few filing cabinets add up to ~1 GB, and a 
> >sizable library runs into the terabyte range.
> 
> Say what?  That's so if you store it all as bitmaps, but that's a
> pretty dumb way to store most books.

Dumb from the standpoint of storage requirements, but compression 
schemes shift the burden away from storage onto the processor. People
won't abandon paper for electronic media until the latter have
similar capability to present information as well as speed and
flexibility advantages. And 98% of the world's recorded information
is in printed form. Until we have some standard way to convert it
all into compact logical page descriptions, the easiest short-term
solution is to use bitmaps for some things. Yech.

The biggest reason people like paper is because it delivers superb
display quality. Paper won't go away if computers can only deliver
80x25 ASCII. You've got to deliver quality screen typesetting, you
need quality grey-scale or color graphics, and you need to have
display speeds approaching how fast you can flip through pages
in a book. Sure, you can strip a book of all its bells and whistles
and pack it onto a double-sided floppy. And wait around for it to
decompress and show up on the screen. Whatever storage scheme you
use, it has to be fast. That limits how tight you can pack it.

I'm writing a book using LaTeX. I have about 1MB of ASCII source, and
the book isn't all that long (about 300 pages typeset, and I'm not
done yet). I have plenty of line drawings, but no photos. Running
LaTeX on the source takes well over an hour on a Compaq 386. The .dvi
file is larger than the source, roughly double. To make the book
pleasantly readable anytime soon, the user needs the .dvi file. To
exploit the searching capabilities of the computer, the user will want
a full-text index, which will require 2-3MB. So 1MB of ASCII source
generates 4-5 times as much storage requirement to deliver the
information usefully. You probably could not compress this by more
than half without paying an intolerable speed penalty.

We can haggle over how we are going to store and present text, but
graphics are the real problem. The need for graphics is not trivial,
else we would have long ago tossed out our magazines and books
and been happy with our character displays. Furthermore, the density
of graphics in books reflects not only the need for them, but the
costs and hassles involved in producing and reproducing them.

If I had photos in my book, the storage requirements would be enormously
greater. To accurately digitize a fine color print requires at least
4k by 4k by 24 bits. That's ~4 x 10^8 bits, or 50 MB. Say we have a 
fast compression scheme that cuts it down by a factor of 50. You still
only get ~500 quality images on a CD-ROM. 

With suitable prestidigitation 50 MB
on a printed page doesn't have to mean 50 MB on the disk. But the
bottom line is the paper piles on my desk can show me those nominal
50 MB, and that's what I want to see, and quickly.

We can always bring up sound and video if we find ourselves with
more storage than we know what to do with. Static descriptions on
paper do not really do justice to many fields of study. Especially
skill-oriented domains such as medicine, arts, laboratory chemistry,
etc. The information content of my paper piles is only a crude
lower bound on how much information I could profitably use if
I could get it quickly and cheaply enough.

About library size: the on-line catalog at my university lists
over 800,000 volumes, and I find what I am looking for about half the
time. Obviously I won't live long enough to read 800,000 volumes, but
an intellectual community needs fast access to a resource at least as large.
Unless I had a very well-defined job that was about to get automated,
I couldn't imagine being able to do very well with only the books I
can personally afford to buy and store in paper form.

Dan Mocsny