Re: Software Quality & Fault Measurement

[Peter Amey <peter.amey@praxis-cs.co.uk>]

Michael Kelly wrote:
> I'm trying to locate information on quality measurement studies that have 
> been done on projects using C, C++, Ada, Java, and Perl. I'm particularly 
> interested in materials regarding fault measurements. Any pointers would 
> be greatly appreciated.
> 
> If these are not the proper newsgroups to post this to, please disregard.
> 

Although it wasn't primarily a comparative study, some useful evidence 
emerged from the Lockheed C130J programme and, in particular, the 
certification effort sponsored by the UK MoD on that aircraft.  Relevant 
papers from Crosstalk Journal include:

"Correctness by Construction, Better Can Also be Cheaper"
http://www.sparkada.com/downloads/Mar2002Amey.pdf

and

"Software Static Code Analysis Lessons Learned" by Andy German, QinetiQ 
Boscombe Down. DoD CrossTalk Journal, November 2003
http://www.stsc.hill.af.mil/crosstalk/2003/11/index.html

Executive summary:  C130J (Hercules II) code, already cleared to DO178 
levels A or B, in a variety of languages from a variety of vendors, was 
reviewed and examined by an MoD-appointed contractor.  Residual error 
rates for Ada were about an order of magnitude smaller that for C-based 
languages.  SPARK had significantly lower error rates than Ada.

Peter