Re: more re: If Ada isn't the problem, what is?

Re: more re: If Ada isn't the problem, what is?

[Wes Groleau X7574]

In article <9304122053.AA10849@ajpo.sei.cmu.edu> SAHARBAUGH@ROO.FIT.EDU writes:
>.............. in a General Accounting Office report.  I have 
>...........
>The longest list is for "lack of well-defined requirements".  I'm not surprize
d,
>are you?

No, I'm not.  And that's obviously NOT Ada's fault.  A related problem not in
the list (or I missed it) is:

"Specs written by semi-FORTRAN-educated engineers who don't know the difference
between requirements and implementation."

Which means that programmers have to try to persuade management that "searching
a monstrous array of characters for a particular substring and using that
substrings location in a computed GOTO" is not really a requirement.

Re: more re: If Ada isn't the problem, what is?

[cis.ohio-state.edu!zaphod.mps.ohio-state.edu!sdd.hp.com!portal!cup.portal]

>>   No, I'm not.  And that's obviously NOT Ada's fault.  A related 
>>	problem not in>   the list (or I missed it) is:
>>
>>   "Specs written by semi-FORTRAN-educated engineers who don't 
>>	know the difference  between requirements and implementation."
>>
>>   Which means that programmers have to try to persuade management 
>>	Ethat "searching  a monstrous array of characters for a particular
>>	substring and using that substrings location in a computed GOTO" 
>>	is not really a requirement.
>
>
>
>I have actually had a requirements document stating that a variable must
>be at a specific memory location.  It did not matter if the variable 
>was even used or not; it had to be at that location.

I like writing software problem reports to systems saying things like:
"remove implementation details from requirement 1.3.34.5"  :-)

Generally they seem to come back with something like "what implementation
detail" and I end up rewriting it FOR them to show them how to not
specify implementation.

Example:

	Was something like:

		Test memory in with incrementing address pattern.

	Becomes something like:

		Test memory with an address pattern calculated to
		exercise all bank select lines and address lines
		on every memory device in all memory banks.

This usually gets a response something along the lines of "oh, that was
what I intended to say in that requirement anyway. I didn't realize
the incrementing pattern was implementation detail" and they go off
an use my example as their requirement change.

                                        R. Tim Coslet

Usenet: R_Tim_Coslet@cup.portal.com
        technology, n.  domesticated natural phenomena