(no subject)

Peter G Capek capek@us.ibm.com
Sat, 27 Oct 2001 17:14:01 -0400


Russell Turpin wrote:

> You would be surprised at the amount of commercial code
> where everything is assumed to work, and the most sophisticated
> response to component failure is to print "foo failed" before
> dying itself.
>
> Or maybe you wouldn't be.

Makes me wonder why subsystems and run-time libs don't have a mode intended
to be used for
testing where, with some probability, they return indicating an error which
could in actuality
be returned from that same call.   This would be very useful for testing
error messages, retry code,
and other code which is difficult to exercise normally, whence the above
comment.   Of course, it may
be possible for the errors returnable to a particular call to evolve over
time, but we'd still be ahead
of where we are today.   I fear this effect is going to be an increasing
problem with web services, where
the two (co)operating parties are under independent development and
administration.

          Peter Capek


Peter G. Capek
IBM Thomas J. Watson Research Center
Yorktown Heights, NY   10598-0218
(+1 914) 945-1250       IBM Tieline: 8-862-1250            Fax:  X 4426