Re: UMLish modellers?
Re: UMLish modellers?
- Subject: Re: UMLish modellers?
- From: Marco Scheurer <email@hidden>
- Date: Sun, 9 Sep 2001 15:59:40 +0200
On Sunday, September 9, 2001, at 02:08 AM, Erik M. Buck wrote:
In my experience, UML modelers are just an expensive time consumer
intended
only to make in experienced non-programmer managers feel like they are
managing something.
Yes...
The only successful large software that I have seen
succeed has been iteratively developed. Start with a kernel of
functionality and keep adding layers, re-evaluate, test, redesign,
repeat
until done. The thing about that approach is that managers hate it.
Yes...
They
can not plan it. Sometimes work has to be thrown out. They can not
say how
many iterations are needed before completion. They can not say when the
project will be complete. They can not say how much the project will
cost.
I think it is possible to plan and budget evolutionary projects. You can
always put a limit on the number of iterations. The plan, which is also
iteratively revised based on user feedback, can simply be, for each
step: do the most important thing. In the end, you'll have a working
system with the most important features implemented, working, and
reflecting the investment made. We start our projects by asking our
customers: "if we had only one week, what is the thing that would help
you the most".
Which is very different from doing the easy stuff first, and as you
said, doing "design" is easy...
The final design never matches the diagrams which means
that the diagrams are not even good documentation. The basic and
essential
problem with modelers is that they do not help with the hard part of the
problem of software development.
Diagrams do not crash, but they do not help the user either. And even if
they were good documentation, UML is not a language you can use to
communicate with the customer. A working program, yes.
Iterations are the price of experimental software development. Many
times,
we just have to experiment and try different things in order to find a
good
solution. A good large application is like a scientific breakthrough.
It
is hard to have a breakthrough on a schedule.
If the customer can not specify what they want, we have to experiment to
discover it.
Yes, and even if they could, it is likely to change during the months it
takes to do all these UML diagrams...
Marco Scheurer
Sen:te, Lausanne, Switzerland
http://www.sente.ch