Re: nancy leveson's book and paper reference

[Miles Parker]

>On Tue, 2 May 2000 address@hidden wrote:
>so the problem is lack of independence. If two versions would be
>independent, output agreement would be sufficient, right?
>
>Of course, true independence can't be achieved, but it is the goal.
>
>And I wouldn't know of anything better than comparing two versions that
>are as independent as possible.

Please someone correct or confirm, but isn't the point that two completely 
independent implementations _will_ tend to recreate errors in the same places, 
thus overturning the assumption that multiple versions of a model validate the 
model? This conclusion challenges some assumptions, but makes a lot of sense to 
me.

As usual, the truth is probably somewhere in the middle. [And this is a problem 
with many absolutist SE approcahes, but don't get me started on that..] It is 
an obvious benefit to have multiple independent 'eyes' on a problem, and I 
think part of the assumption is that they will not bring the same hidden biases 
or blindspots to an implementation. At the same time, people are a) exposed to 
many of the same bisaes and blindspots and b) work within the same broad 
syntactic paradigm, not to say platform, language, and framework, so it is 
likely that n-versions might make the same mistakes in the same areas, and that 
'correct' solutions (esp. if novel) may well be outriders.

Hmm, or perhaps I should just read the book instead of BS'ing about it...:-)

-Miles





>> hi, all -
>> 
>> nancy leveson's book is
>>      safeware: system safety in the computer age
>>      addison-wesley 1995
>> 
>> her paper on n-version programming is
>>      j.c.knight and n.g.leveson
>>      an experimental evaluation of the assumption of independence
>>              in multi-version programming
>>      ieee transactions on software engineering, volume se-12,
>>              number 1, pp. 96-109, january 1986
>> 
>> there are numerous others - see http://sunnyday.mit.edu/ 
>> 
>> more later,
>> cal
>
>
>
>                  ==================================
>   Swarm-Modelling is for discussion of Simulation and Modelling techniques
>   esp. using Swarm.  For list administration needs (esp. [un]subscribing),
>   please send a message to <address@hidden> with "help" in the
>   body of the message.
>                  ==================================


Miles T. Parker
Software Engineer
The Brookings Institution  1775 Mass. Ave. NW  Washington, DC  20036
http://www.brook.edu/es/dynamics/models/ascape
mailto:address@hidden  voice 202.797.6136  fax 202.797.6319



                  ==================================
   Swarm-Modelling is for discussion of Simulation and Modelling techniques
   esp. using Swarm.  For list administration needs (esp. [un]subscribing),
   please send a message to <address@hidden> with "help" in the
   body of the message.
                  ==================================