Path: utzoo!utgpu!water!watmath!clyde!rutgers!ames!pasteur!ucbvax!JVNCA.CSC.ORG!salzman
From: salzman@JVNCA.CSC.ORG (David Salzman)
Newsgroups: comp.society.futures
Subject: Re:  CD ROMs
Message-ID: <8802011427.AA06366@jvnca.csc.org>
Date: 1 Feb 88 14:27:54 GMT
Sender: daemon@ucbvax.BERKELEY.EDU
Organization: The Internet
Lines: 26

There are two reasons why Read-Only mass storage systems will be important,
assuming that they meet a price/performance window and become reliable.
(1) Data sets have strongly bi-modal endurance.  If you were to plot the
"lifetime" of every block written to disk, you would not observe a decaying
exponential.  Instead, you would see a fast decay with a characteristic
time on the order of a day, quickly leveling out.  A rule of thumb is that
if you haven't changed a data set within a week, you'll keep it forever.
If a CD ROM is not worse than a Winchester in price, then its greater
stability makes it attractive for the role of a long term storage archive.
(2) The optical CD ROM you know and love from the music world will reach
fundamental physical limits (i.e. wavelength of light) at something like
1-10 bits per square micron (1E7 bits/sq.inch), if you assume that the
lasers used to write and read are near the visible spectrum.  Unless there
are electron-optical systems available, limited to 1E3-1E4 bits per square
micron, fundamentally different technologies will have to be used.  The
non-2D systems with random access include systems with more states than
just 0 or 1, and states with solid-state logic, which are both exotic.
Instead, 2D non-random access systems (e.g. Digital Audio Tape) are likely
to predominate.

In other words, a more realistic comparison than CD ROM v. Winchester is
CD ROM v. DAT.  Do you want random access, but read-only (or at best 
write-once), or read-write, but serial?

 - David Salzman
   John von Neumann National Supercomputer Center