On 12/2/2013 6:13 PM, Ihe Onwuka wrote:
> That would be Gall's 3rd System Axiom.
>
> http://www.laetusinpraesens.org/docs/systfail.php
>
> Even with a good clear set of requirements, analysts and designers
> dilligent and good enough not to specify/implement an over-engineered
> solution are thin on the ground.
Anyone who has been in engineering or manufacturing for any significant
time should know that things ... systems ... whatever that has any
complexity ... hardly ever work as expected right off the bat. Things
have to be tweaked, fixed, redesigned, reworked before they work right.
A new ship, for example, will likely have thousands if not tens of
thousands of item on its punch list when first launched.
That's true about software, too. Anyone who thinks a software system is
going to work right at first is a fool, there's no way around it. But
the contractual and managerial means that would provide for the
evolution to a correctly working system - and these means have to work
all during the project, not just near the end - we hardly ever have them.
This would be true even if the requirements were fully known and
properly worked out from the start, and the development team were
totally capable and understood how to use the tools provided. But we
know that's hardly ever the case, too.
When legal and political matters are involved, they make it much harder
because they can't be flexible where flexibility is needed.
As I think about this, it seems to me that what is needed is research
into new methods of contracting for and managing software systems. It's
not that a new methodology will solve everything. It's not that a new
modeling tool or a new programming language will solve everything. It's
not even that better programmers will solve everything. The
social/managerial/contractual base is completely inadequate.
Hopefully what emerges will be simple. Enterprise Architecture was a
apparently good idea that grew out of control into an unmanageably
complex monster. And yet at its simplified core, it still can have value.
