Cherry-picking

[Mike Meyer]

On Wed, 9 Jun 2010 11:34:34 -0500
"David Dyer-Bennet" <dd-b at dd-b.net> wrote:

> 
> On Tue, June 8, 2010 14:46, Mark A. Flacy wrote:
> > On Tue, 2010-06-08 at 13:36 -0500, David Dyer-Bennet wrote:
> >
> >> You can't "guarantee" (in any meaningful sense) both completion date and
> >> feature list.  "Estimates" are just that; they're not guarantees, and
> >> trying to force engineers to treat them as guarantees is basically
> >> futile
> >> (and leads to death marches and really shitty software).
> >
> > You have got to be joking.
> 
> Not joking.
> 
> > If you can't do that, you aren't doing engineering.
> 
> Some truth to that; what's called "software engineering" is pretty
> marginal as an engineering discipline.  What we do would be "research and
> development" in most other fields, not engineering.

We also deal with issues that engineers in other fields wouldn't dream
of doing. When they build a bridge, they expect the maximum traffic
load estimates to be relatively constant over the expected lifetime of
the bridge; we're asked to build "scalable systems" that can be built
for a cost suitable for one load, then be grown to handle orders of
magnitude more load at a cost that's only marginally greater than
having built it for that load to begin with. Etc.

> No doubt there are places that deal with simple enough software problems
> in stable enough software environments that they can do actual
> engineering, but I've never actually seen one.

From what I've heard, NASA qualifies. A lot of what they do could be
adopted elsewhere (debugging the process being at the top of the
list), but they also have a completely fixed environment and goals for
a project.

> And if schedule is important enough, you bias your estimate; work towards
> "no later than" rather than "best estimate".

That, of course, is part of doing engineering vs. doing r&d. You give
worst case estimates instead of expected case, so you can look like a
genius when you finish ahead of schedule.

     <mike
-- 
Mike Meyer <mwm at mired.org>		http://www.mired.org/consulting.html
Independent Network/Unix/Perforce consultant, email for more information.

O< ascii ribbon campaign - stop html mail - www.asciiribbon.org