[
Lists Home |
Date Index |
Thread Index
]
>On Tue, 2005-01-04 at 16:41 +0000, Kirkham, Pete (UK) wrote:
>> All these documents require inspection by
>>multiple human experts prior to issue, and are
>>text or graphics rather than data, so type
>>systems do not figure.
>>
>> Since there are occasional semantic errors (m
>>may be miles or metres- the authority rings up
>>the author and it is expanded, currently the
>>different AWFLs for different planes need
>>compiling by hand rather than automatically via
>>a query off a database) I've thought about
>>using a controlled English that is machine
>>parsable (rather than presenting the users with
>>XML or RDF), but the gain for such a system is
>>not likely to exceed the cost- most of the
>>checks worth anything are in the heads of the
>>grey-haired engineers, not any type system.
>>I've had a lot of experience with trying to get
>>Z, graph based argumentation and Goal
>>Structuring Notations into such systems for
>>safety case management, but without much
>>success. Basically the people who are experts
>>in safety don't trust automation when there's
>>an extant reliable alternative.
>
>Exactly. Real life error conditions almost never fall along the neat
>lines that strongly, statically typed language designers draw. I don't
>think I can recall having *ever* had a program fail because someone
>passed a float to a routine that expected an int. Most errors by most
>competent programmers (IMO) are of much sterner stuff. And the problems
>with most strongly, statically typed languages is that their strong
>typing decreases expressiveness of axioms to the extent that it's not
>even easy to arrange to catch the real sorts of errors that can occur
>(although, luckily the recently revived test-first philosophy is helping
>a lot with that).
>
>I have never believed the notion that strong static typing increases
>safety, and that's for general-purpose languages. In the case of XML
>applications, I go further to consider such ideas of type safety fairly
>ludicrous.
You might consider studying the works of
Mandelbrot, Lorenz, Spiegel, Winfree, Laplace,
Poincaré Neumann, et al with respect to Chaos
Theory. For example, Mandelbrot was hired to make
a fail-safe system for the telephone company.
Unfortunately, as he proved, that's not an
option, for all systems become chaotic.
Regardless of the design, redundancy, quality of
equipment, or testing of the software, they will
ultimately fail.
The only thing one can do is design systems that
expect and deal with that possibility.
tedd
--
--------------------------------------------------------------------------------
http://sperling.com/
|
