[SystemSafety] Fault, Failure and Reliability Again (short)
Derek M Jones
derek at knosof.co.uk
Wed Mar 4 14:33:37 CET 2015
On 04/03/2015 12:54, C. Michael Holloway wrote:
> Some
> have claimed that the work conducted in the field has not yet borne any
> healthy fruit. Some, myself included, doubt it ever will.
I am more optimistic about fruit being there to find, but given a
major change of research direction is needed less optimistic that
it will be found anytime soon.
The problem is that researchers are acting like drinks searching for
lost keys under the streetlight, because they can see there.
Almost all existing papers on fault result ignore the input
distribution. It is easy to obtain lots of opens source fault
data from the enumerable bug report databases and the code is
available; getting information n the input distribution is very hard.
The situation is not helped by the fact that machine learning
is all the rage these days and people are feeding anything they
can get their hands on into learners and publishing the results.
Taking the time to clean the data before using it is a lot of work.
http://shape-of-code.coding-guidelines.com/2013/06/02/data-cleaning-the-next-step-in-empirical-software-engineering/
Empirical software engineering is still very young. We
should not prejudge a field that is barely 10 years old:
http://shape-of-code.coding-guidelines.com/2011/03/31/empirical-software-engineering-is-five-years-old/
--
Derek M. Jones Software analysis
tel: +44 (0)1252 520667 blog:shape-of-code.coding-guidelines.com
More information about the systemsafety
mailing list