BKMTHSTM.RVW   980201

"Mother of Storms", John Barnes, 1994, 0-812-53345-3, U$5.99/C$6.99
%A   John Barnes
%C   175 Fifth Ave., New York, NY   10010
%D   1994
%G   0-812-53345-3
%I   St. Martin's Press/Tor Books/Tom Doherty Assoc.
%O   U$5.99/C$6.99 212-674-5151 fax 800-288-2131 pnh@tor.com
%O   www.tor.com www.stmartins.com
%P   560 p.
%T   "Mother of Storms"

Of storms and meteorology I know very little.  Weather in Vancouver is
frontal, and that makes for a rather boring, though not entirely
predictable, progression.  The description of global warming and storm
formation in the book seems plausible, and not inconsistent with what
I do know.

What I do know, of course, is computers, communications, and viruses. 
These play a large role in the book, pretty much as large as the
hurricanes themselves.

Let's start with the viruses.  First off, the entire computing
infrastructure of the book appears to be very similar to Fred Cohen's
theorized viral computing environment.  A far cry from the simplistic
code fragments of today, the programs of the book's time are all, to
some extent, viral and replicable.  They also seem to use a lot of
genetic programming, modifying themselves sometimes in light of a
superior implementation, and sometimes at random.  However, perhaps
because of a lack of comfort on the part of the author, the technical
population of the book seems to have lost control of the technology. 
The programs are left to propagate, replicate, and develop as they
will.  One character allows as how the programs are now so complex
that nobody can understand them.  For me, this sounds a little bit too
much like the "computers work by magic" attitude that absolves people
from responsibility for learning and thought.

Secondly, there are the specifically viral programs of the book, the
datarodents.  These are targeted information gathering applications,
which roam the networks, populating nodes and routers where the chance
of finding the desired information seems high.  When a datum is found,
a datarodent will replicate itself to a) stay at its post, b) find a
path back to the originator to report the datum and/or c) trace back
the route of the datum to its source, there possibly to replicate
further.  This behaviour was fairly standard theorizing in the early
days of computer virus research.  (Due to the profligacy of basement
variant makers, and their total lack of imagination, virus research is
not as much fun as it once was.)

The activity of the datarodents, however, appears to be inconsistent. 
On the one hand, nonentity individuals seem to be able to find almost
any information in almost no time at all.  Every conversation can be
overheard and reported.  The entire world lives in a fishbowl in which
nobody has ever heard of encryption.  (Or, if they have, the author
doesn't really understand it.  On the other hand, network replicable
programs would make terrific encryption breakers.  On yet the third
hand, self-optimising programs would not be a particularly good bet,
since speed of operation is quite measurable, and would presumably be
selected for, while strength of encryption is very difficult to
assess.  You might end up with a very fast, but completely insecure,
implementation.  But enough digression.)  Except that not all
conversations *are* overheard.  Important information goes unfound. 
Lost datarodents (of which there must be more than a few) for a given
project cannot be found and the project reconstructed.

Then we come to the major event: the melding of man and machine.  Now,
I don't have too much of a problem with the "optimization" of the
programs.  As mentioned, this is a logical and reasonable followon to
the current work in genetic programming.  I don't really even have a
problem with cross-platform programming.  This exists, in some guises,
in current systems.  There are, for example, the fat binaries for the
Macintosh, which hold code for both the 68xxx and 88xxx processors,
and also cross-platform environments such as Java and the MS Word and
other VBA compliant macros.  I can even grant, rather dimly, some
connection between neural net computing, and our own neural activity. 
(The author is, by the way, to be commended for being the first writer
I can recall to understand that the brain changes, under experience,
and that this would have an impact on direct neural communications.)

The actual crossover of computing programs to the human brain is,
though, a bit too much like the old joke of the masses of equations on
the left side of the blackboard, the desired result on the right side
of the blackboard, and the cloud in the middle saying "and now a
miracle occurs."  As the punchline goes, I think you need to be a bit
more specific in step two.

copyright Robert M. Slade, 1998   BKMTHSTM.RVW   980201