Re: Testability and normative requirements

Sometimes making a requirement testable, is a matter of phrasing it in a 
way that is testable.  however, there are some cases where them may be a 
requirement that can't be tested - for example, an implementation shall not 
limit the size of the data file (I think there was something like this in 
CGM - ISO 8632).  Also, what about security requirements - are they always 
testable?  For example, the system shall detect viruses.  Can this be 
tested - since we can only test for those viruses that are known and can't 
prove that the system can detect a virus that isn't know.
So, the question then becomes, is the requirement really needed, since it 
can't be tested?   Is it really a requirement?
I don't advocate untestable requirements, but in reality, I think that it 
may be necessary to include some in a specification.

-lynne



At 09:44 AM 12/21/2005 +0900, Karl Dubost wrote:


>Le 05-12-21 à 02:22, Dominique Hazael-Massieux a écrit :
>>Is there a logical contradiction behind the idea of a normative
>>requirement that would not be testable? I don't think there is, but
>>would be interested to hear what others think about it.
>
>Could you give a practical example of what you would consider as a
>non-testable normative requirement?
>
>Often when I had to deal with this it's because we don't test the
>right thing. My favourite example for that is coming from nuclear
>physics. You can't test the position of a particle, but you can test
>the probability of its position. The first one is non-testable by
>virtue of the physics law (normative requirement) but it doesn't mean
>you can't achieve a test which verifies the law. There's a
>probability that  this particle is here or not. Does the test verify
>this probability?
>
>
>--
>Karl Dubost - http://www.w3.org/People/karl/
>W3C Conformance Manager, QA Activity Lead
>*** Be Strict To Be Cool ***
>
>
>
>
>

Received on Wednesday, 21 December 2005 01:05:14 UTC