[>From Le Monde, dated 8 June 1996, i.e. published on the 7th; on-line
  edition at http://www.lemonde.fr. Extracted and translated by BM.
  (Although ellipses are not marked, I have considerably abbreviated the
  text and removed some of the anthropomorphic comments, e.g. "the machine's
  brain" and the like. Comments in square brackets [] by BM.)]

THE MYSTERIES OF ARIANE'S CHAMBER, by Jean-Francois Augereau

Who [sic] caused the in-flight explosion of Ariane-5 on Tuesday, June 4?
After more than forty-eight hours of preliminary investigations, "witnesses"
are starting to talk. The propulsion system, which could have been
suspicious because of its novelty, has been cleared. The likely culprits are
elsewhere, "in the software or the hardware", that is to say the
computer-related parts. Only five of them are left, gathered in one "closed
room". [???]

According to Daniel Mugnier, head of the Launchers ("lanceurs") division at
the CNES (National Center for Aerospace Studies), the inquiry is focusing on
the "electrical and software system" which allows the various elements of
Ariane-5 to talk to each other. The launcher is loaded with sensors which
constantly monitor its moves and accelerations.

Our first suspect is an Inertial Reference System (IRS)*, the balancing
center of the launcher. The IRS, or its mate, is in charge of using these
data to compute the launcher's exact position, speed and acceleration. But
at this stage of the inquiry it seems that the sensors themselves have been
exonerated.

There is no alibi, however, for the IRS. Doubts remain, even though the
on-board computer and the backup unit show a record of having received
[litt. "claim to have received"] the same information. How could they have
failed at the same time and in the same way?

Hence the questions about the behavior of on-board computers. According to
Daniel Mugnier, "they ``claim'' to have received abnormal information from
the IRS. Whom [sic] should we believe? Daniel Mugnier is reluctant to
incriminate that component [i.e. the computers?].  Same thing with another
component, the "1553 bus". It is a kind of information highway [??!!]; all
navigation commands go through it. According to one of the investigators,
"it is a proven system, which has been used for a long time on all NATO
fighter planes".

This leaves two other suspects: the in-flight software program and the
coder.  Does the program, made of long lines [???] of computer writing,
include a "bug" or a fault? Did the converter**, which translates the
sensors' analog language into the computers' digital language, stutter? One
cannot exclude the possibility that the computer is denouncing errors that
it itself created.

The investigation continues. The report should be turned in by July 15.

[Notes:
    * I have translated "Centrale Inertielle (SRI)" by "Inertial
    Reference System (IRS)". I found the acronym in Jane's Defence
    Glossary at http://www.thomson.com/hanes/janesgloss. I don't
    believe it's directly connected to the Internal Revenue System.

    ** I used "converter" for the analog-to-digital "codeur".]

Bertrand Meyer
ISE Inc., Santa Barbara, <bertrand@eiffel.com>, http://www.eiffel.com