Re: Ada Program Management Questions

Re: Ada Program Management Questions

[David Emery]

In article <4id0t8$rm5@news7.erols.com>, dmurphy@mail.erols.com (Dennis
Murphy) wrote:

> I am seeking some information on Ada, I have three questions:
> 
> 1)  What is the industry standard for producing a single SLOC (in dollars)?
> 
> 2)  What is the industry average cost for a Software Engineer, per hour to 
>      produce code in Ada?
> 
> 3)  What is the average SLOC per hour for an experienced programmer?
> 
> Any responses to these questions will be greatly appreciated.  If anyone 
> knows of any publications I can review to find these answers, please 
> post these messages also.  Please post them to this newsgroup and forward a 
> copy to my email address at work, as I do not have access to this
newsgroup at 
> work.  Below is my email address.  Thank you.
> 
> Jack Murphy
> jmurphy@cc.atinc.com

My experience is that 
   1.  Taken across a very wide sample, people seem to be able to
      write/debug betwen 100 and 200 SLOC per month.  This has been the
      "going rate" for about 20 years, or since we started measuring code
      in SLOC/unit time.  
   2.  There is a very wide variation in the productivity of individuals.  Some
      individuals produce 10 times more code than others.
   3.  Ada doesn't change the rate of production much.  But it does change
      the allocation of effort. You spend more time designing, and less time
      coding and particularly debugging.
   4.  Ada starts to provide "a win" when integrating.  In particular, Ada 
      systems in the 10k-100k SLOC range seem to go together quite smoothly.
      Above 100k you have to have additional tool and procedure support.

            dave

Re: Ada Program Management Questions

[Michael F Brenner]

Although it is more popular in the literature to measure code volume
by SLOC, there is no meaningful correlation between SLOC and effort
at Software or Firmware MAintenance Time: it takes about the same amount
of effort to change 10,000 lines of code as it does to change 10 lines of
code, because most of the effort is in analyzing the impact of change;
alternatives to SLOC that Are correlated to maintenance effort are:
(1) the amount of time it took for That organization to maintain code
of That type the last few times and (2) the number of function points
being changed. Unfortunately, the only textbook that most people seem
to have on their desk is a famous expositor of SLOC's. Be careful to
avoid negligence when measuring software maintenance effort by SLOCs,
and make sure you can personally justify some kind of correlation in 
your particular case between SLOC and effort. This is an example of 
measuring what is easy to measure, rather than what is useful. For the
year 2000 problem, for example, a better estimate than SLOC for changing
the two-digit years to Julian dates or to four-digit years is $250,000
per chip that is opened up, rather than any SLOC count. Just because your
boss asks for SLOCs does not mean that you are telling the truth, if you
give her SLOCs.