Re: [xml-dev] Designing an experiment to gather evidence on approachesto

[Michael Kay <mike@saxonica.com>]

On 29/12/2011 19:04, David Lee wrote:
> This is still a curious subject to me
> Can we compare/learn from other engineering fields?
Back in my ICL days we tried to develop a discipline that recognized 
five key qualities of an IT system:

* performance
* availability
* security
* usability
* potential for change

Performance is easy to quantify and measure, and we do it all the time.

Availability, security, and usability are expensive to measure 
scientifically, and most projects take short cuts. But if they are 
important enough, we know how to do it.

Potential for change accounts for an enormous proportion of the design 
effort we put into IT systems: all the stuff about modularity, 
interfaces, high level languages, abstraction, encapsulation, and the 
rest. And yet we have no idea how to measure it at all. We invest a lot 
in potential for change, and we have no idea what return we are getting 
on our investment.

I think that in this respect software engineering is different from 
other branches of engineering. It is soft, so potential for change is 
important. No-one designs bridges with a view to changing the bridge 
when technology changes or requirements change, except perhaps in 
certain very well-controlled ways, like designing in the ability to add 
an extra lane to a motorway.

Michael Kay
Saxonica