A review of the CD-ROM version of the
 5th edition of Table of Integrals and
 Series by Gradshetyn and Ryzhik,
 as published by Academic Press
  in collaboration with Lightbinders,
  and based on the Dynatext software technology by EBT Inc.

reviewed by Richard Fateman 10/21/96 for SIGSAM Bulletin

publisher's suggested retail price: $79.95 CD-ROM ISBN: 0122947568
publisher's suggested retail price: $63.00 book ISBN: 012294755X


The object under review is a CD-ROM publication, newly produced in
July 1996, but based on the classic table of integrals, originally
published in Russian, and now in its fifth edition.

In this case the text and the mathematics have been newly typeset but
into an electronic master copy.  From this master copy, one can view a
page with a typeset appearance but, at least in principle, one can
extract other useful electronic forms.

In particular, an effort has been made to provide a {\TeX} version of
most of the 20,000 or so formulas in this edition.  \TeX, a
typesetting language designed and implemented by computer scientist
Donald E. Knuth is widely used for journal publications by the
American Mathematical Society and other publishers.

The identical CD-ROM can be run on OS 4.1.3 or Solaris for Sun
workstations; on Silicon Graphics IRIX 5, on Macintosh 7.1 or later,
and on Windows 95.  We tested it on SunOS Release 4.1.3_U1 using X11R6
running on a Sparc 1+ processor, and also on Windows NT4.0, running on
a very fast (Pentium Pro) processor.

The CD-ROM has two components: a collection of data constituting the
integral table, and a program which presents the information on
request to the user. There is also an installation program that copies
what needs to be on your hard disk off the CD-ROM.  The data is
essentially that of the printed version, with some errors corrected
from previous editions, and presumably with new errors introduced.

The major question facing a purchaser of this new edition is: Should
this CD-ROM be purchased instead of the ordinary paper (OP) text?
Assuming you are equipped to read the CD-ROM, I see a few possible
reasons to favor it over the OP.

1. If you have the (unusual) habit of picking integrals
out of this table and typesetting them into your own papers.

2. If you typically have several simultaneous users of this book, each
of whom has access to the same computer either locally or by
communication link.  This CD-ROM comes with a license that asks that
you limit usage to 5 users at a time.

3. If you wish to save shelf space, or shipping weight.
Unfortunately, although the CD-ROM itself weighs much less than the
book, the CD-ROM reader plus computer weighs considerably more. If you
own many other items in your library on CD-ROMs, the weight advantage
would tilt back toward the computer version.

4. You are a hard-core techie who prefers electronic over paper
versions.


The reasons I expect most users to prefer the OP version to this one
are compelling.

1. Faster acess, assuming you have it on a bookshelf nearby. 
2. Better readability. 
3. Low (i.e. zero) power consumption.
4. High portability.
5. Relative permanence (not that CD-ROMs fade, but who knows
whether readers for this CD-ROM format 
will remain common as higher-density, smaller, faster,
etc. technology appears.)

and perhaps

5. Cost. The OP costs less than the CD-ROM version, according to
the publisher's suggested pricing.

Although Dynatext allows for search, it does not match integrals.
It is really effective only on the English words in the introduction
and in sub-chapter headings, not on the mathematics.

In addition to this overall comparison, we have many quibbles with the
CD-ROM, and some serious issues.  We deal with the quibbles more or less
chronologically as we encountered them in trying to use the CD-ROM,
first on the X-window Sun system.

For a start, the CD-ROM arrives without instructions. One must know
the correct incantation to type to a Sun workstation in order to read
the installation directions (at which point they are redundant).
Even so, the instructions were wrong or inadequate regarding the
.ebtrc file, and I had to find a local expert to find out about the
XKEYSYMDB environment variable.  For Windows NT, the installation
is fortunately uneventful: one clicks on the setup icon.

Next, the vendor avoids telling you that the contents of this CD-ROM
can be copied to a hard disk, and (in my opinion) must be so copied to
be economical.  If one were to use it as (apparently) anticipated by
the vendor, one would either need to devote a CD-reader computer
peripheral full-time to this disk, effectively adding a few hundred
dollars to the package, or one would have to insert this disk by hand
into the computer each time it is needed.  This would make a very slow
operation even slower, and would require that the user physically be
near the computer (rather than, say, using the computer via a
network.)

Computers are supposed to be fast.  How fast is this?  Having loaded
the data to a hard disk saves time, but even so, I found it painfully
slow on the Sun. Starting up the program on a Sparc 1+ (admittedly not
the fastest machine these days, but not a total slug), takes 20
seconds.  I would complain less if it were doing something useful, but
it appears that 15 seconds of that is to load the superfluous
title-page color image.  The image is of the rainbow-hued CD-ROM you
had in your hand a few moments ago.  And incidentally, it appears to
use up much of the workstation's default colormap.

On the Windows NT machine, a 200 megahertz Pentium Pro system, the
startup time was negligible (a second or so), even using the CD-ROM
(an 8X reader).  In order to access the fonts needed for viewing, it
is apparently necessary to reboot the system after installation.  The
Windows installation instructions also seems to expect that you will
devote your CD-ROM reader full-time to this material, but some
fiddling with the file dynatext.ini and moving the AP directory to my
hard disk enabled me to run with the CD-ROM removed.

What about the presentation?

The main technique used in the CD-ROM provides a dual view of the
material: by default on the left of the main display is a
table-of-contents "TOC" in outline form. If one selects the top-level
outline headings by clicking, they expand to sub-headings. Opposite
the TOC on the right is a frame consisting of a glimpse into the
full-text form of the document corresponding to the selected part of
the TOC.

Unfortunately for some of the GR material on some displays, the full
text form of equations is too wide to see on the window, and no way of
panning from left to right is provided. The only way I found of
viewing large expressions was to stretch the window, wipe out or
decrease the part consumed by the TOC by moving the central divider,
and/or decrease the font size of the display.

Also unfortunately, the font size chosen by default (on Sun) is
inconsistent with the given spacing size.  Characters sometimes fall
on top of each other. Characters are missing from the default font,
too.  I found that if I chose to increase the size of the type, it
became more readable, and missing characters were correctly displayed.
However, this exacerbated the problem with large equations, which
would simply become impossible to see in their entirety. These
problems were not apparent on the Windows system, however.

The contrast was startling. a few miscalculated defaults on the Sun
system made reading the material quite painful. Readjusting the
system parameters might repair this problem, but frankly,
one does not expect to need a reference manual just to browse through
a book.

To be fair, GR is not an ordinary book: even the OP version of GR has
what amounts to a user's manual in the introduction. It explains how
the book is organized and provides some general guidance on the
mathematical semantics of the entries.  Of course this same knowledge
is needed by the CD-ROM user, as well as computer guidance.

What else is special? Hard to use annotations, bookmarks, hyperlinks.

How good is the manual and the on-line help? Not so good.  The "About"
instructions has figures that are displayed just a bit too small to
read. Only if it occurs to you to click on the figures, are they
re-displayed larger.  The menu "help" selection tells you that help is
available, but you must play hide and seek to find it.  Here again you
must realize that you must click on figures and tables in the help
text to read them.  Some mechanisms that seem to be available are
unexplained (like Journals). To my dismay the clipboard mechanism does
not work in the Sun/X-windows environment.

There is an annotation/ bookmark mechanism that perhaps has useful
generality, but its complexity is rather daunting compared to the
simple notion of using a bookmark.  Certainly many users of the OP GR
have put a paper tab in at the list of notations. A computer version
of such a tab might be handy, and it may even be possible to set this
up.  Sorting or editing annotations, and keeping track of them either
on a public or private basis, may have a better basis for design than
this document. Controlling notes in a large group of publications all
under a similar controlling program might benefit from such
techniques.

About the package

There are many source-code documents in the bundle of material on the
CD-ROM. At least some of them appear to be dross left over during the
production and manipulation of the content, and presumably need not be
loaded into a computer.

The basic encoding of information is in SGML. For the uninitiated,
this stands for Standard Generalized Markup Language.  And for the
SGML initiates, it does not use a Math DTD. Instead the version used
here makes extensive use of "escapes" to {\TeX}. It is possible to
dump to a file, pieces of the encoding of the book for
examination. Apparently there is another representation of the
formulas in DVI format, an outgrowth of {\TeX} for Device Independent
graphical code, also stored.


There are many ways in which the Dynatext program and the data as
provided on this CD-ROM, could be reformulated to assist the
user---and make the electronic barrier worth overcoming.

The system speed may be acceptably fast only on rather fast machines.
Even so, a fast system my not overcome the advantage of paper, just yet.

My view of how I would use such a program would be as follows: I
expect to be working with a computer algebra system nearby (common
ones include Mathematica, Maple, Derive, Macsyma) in which I have
formulated some problem requiring information from GR.  Contrary to
popular belief, and even advertising claims, these programs do not
make tables like GR obsolete. The program could then look up
integrals on request, making appropriate matches and returning the
answer, with appropriate parametric substitutions and
simplifications.
The user need not view page images at all, although it might
be useful to check these in some cases.

Perhaps future editions of the system would support such a scenario,
recognizing the merits of communicating this information to computer
algebra systems or combined word-processing/mathematical
interfaces such as Scientific Workplace. Such a linkage
would make such a CD-ROM distinctly more valuable. It is an open
question as to whether the
mapping from the \TeX encoding to that suitable for such processing
can be automated.

Richard Fateman
Berkeley CA