Re: meta-specs (was RE: A few things I noticed about w3c's xml-sc hema)

[Francis Norton <francis@redrice.com>]

"Thomas B. Passin" wrote:
> 
> [Bullard, Claude L (Len]
> 
> > I'm wondering if for some projects, it isn't also
> > the best way to start.  In other words, should
> > one create a UML-like design if what is being
> > designed is NOT an OOP, but simply the data
> > exchanged?
> 
> You're getting dangerously close to my favorite design approach - which
> really starts with requirements elicitation - namely, Ian Graham's SOMA
> method(ology).  He advocates considering the system as interacting with
> people (or other computer systems) via a series of (abstract) messages.  The
> messages convey some data and a request, the return provides data or
> confirmation or whatever.  It is the responsibility of various tasks to
> accomplish the golas of the messages.
> 
I'm not familiar with SOMA, but I'd like to agree strongly - my
experience is that the way to get high quality, loosely coupled,
naturally scoped systems is to design from the interfaces in.  That's
why I push the idea of web services that send and receive schema-valid
XML messages.

Francis.