[OT] Evolutionary Project Development Re: [ffl-devel] Statistics

[Ross G. <ro...@sa...>]

This has gone off topic quite a bit, if anyone would like to follow up 
privately please feel free...

Mike Noyes wrote:
> On Fri, 2003-02-28 at 10:22, Ross Gardler wrote:

> If you want to take extreme programming a step further, take a look at:
> 
> Evolution as a project development model
> http://www.mail-archive.com/leaf-devel%40lists.sourceforge.net/msg04541.html
> 

Interestng...

As it happens I have done a great deal of work on Evolutionary software 
development. My approach is to build a "maturity model" that is used to 
guide the project development, that is specifiy a set of expected future 
goal states and identify small steps between them.

This maturity model informs the selection/adaption/creation of best 
practice business models, these in turn inform the 
selection/adaption/creation of best practice software patterns. Test 
harnesses are then built to ensure requirements coverage (as we 
currently understand them).

These patterns and tests are fed into a application builder which 
sources the components required to implement the software. What you end 
up with is either a working system, or a bunch of test failures and a 
partial system. Work on the components and eventially you get a full 
system. furthermore, since you have continuous automated builds when 
someone else enhances a component you may find that your project takes a 
step towartds a higher level of maturity automatically.

Components need not be written in a specific language, however, they do 
need to operate over SOAP, XML-RPC or better still a REST style 
architecture. Of course existing legacy components can be wrapped 
accordingly. This means that existing component projects can easily be 
incorporated into the system and software built this way sees ythe 
benefits of true Open Source component level evolution.

Of course, requirements change - simply change the test harnesses, the 
build system automatically reconfigures your software, or tells you why 
it can't do so. Expected goals also change - simply change the maturity 
model, associated business and software patterns, and the test harnesses 
  for the changed future positions and the build system starts moving 
towards the "new" future.

I am currently unable to publish most of this work due to restrictions 
on publication by some of the involved organisations, however, I am able 
to release the software as Open Source and fully intend to do so just as 
soon as I can find the time, which will be soon. However, I've recently 
relocated to Trinidad and Tobago and am spending most of my time 
enjoying my new life (I just experienced my first Carnival, if you 
people have never been here for Carnival I can highly recomend it).

If you are interested I'll be sure to announce the project here when it 
goes live (Note CSMS will be built in this way, and so all the features 
in there will be available as components in this way).

Ross