Modelling in CS and Physics

[Aiviet Nguyen <aiviet2@yahoo.com>]

Hi guys,

   We have not discussed meta-science since a long time. Today, I would
like to chat a not so meta-scientific but still somewhat
meta-technologic topics: modelling in CS vs the one in Physics.

   Modelling in CS ( more exactly in Sofware Engineering) is still a
very adhoc task. People are talking about the Art of Object-Oriented
Design. There is no unique procedure. The Design depends heavily on the
experience of the architects. There are tools like 
E-R diagram techniques, Shaler-Melosh, Rumbaugh,...Still they are not
derivable from first principles as in the best modelling discipline,
Physics.

   Perhaps, because Physics is an old discipline and physicists have
learned from generation to generation how to become masters in
modelling. But in fact, modelling in Physics is already a procedure.
You just have to choose the first principle and bingo: it's matter of
carrying out the calculation.

    Since Wigner ( the longest lived gentleman physicist), the
principles have been known as symmetries. Symmetries can be observed
from a give domain of knowledge. If your observations, data say that
there are some regularities a careful and methodical analysis on the
data can suggest a symmetry
(exact or approximated one).

   The objects in Physics then can be defined as representations (
multiplets) of the found symmetry group. The quantum numbers are the
eigenvalues of the Casimirs operators... Everything is straightforward.


   A beautiful example is the classification of particles based just on
the Poincare symmetry. From a simple observation of the regularities
which can be observed by even our ancesters in the Stone Age like 
isotropy and homogeneity of space-time, particles come
out to be characterized by mass and spin in a beautiful hierarchy. In
my opinion the glory belongs to Wigner.
Another example is the unitary symmetries of Heisenberg and Gell-Mann.

   Since then, modelling in Physics became a technology. In any Theory
of Everything if you postulate a symmetry, the objects in the modell
are identified. You can decide to consider a part of it in a given
application.

    My question here is whether we can develope a similar procedure in
Sofware Engineering beyond the adhoc E-R diagrams. Is there any
principal reason to prevent us from doing that? Because starting points
of CS and Physics are different? What are they?

    If yes, I would go back to the old question I have raised long ago
about objectification and structural program ( for example, C++-ization
a C program). If the analogue works perfectly, all we can do is to find
a symmetry ( based on data analysis). There should be some program to
take the symmetry group and the C program as inputs and C++ program
together with its objects will come out wonderfully.

    And we can make a lot of money :-)) Reality or just a
meta-technological dream, I don't know for a moment. But it is always
worth thinking. Any creation start from a meta idea like this.

Cheers
Aiviet







    


_________________________________________________________
Do You Yahoo!?
Get your free @yahoo.com address at http://mail.yahoo.com