Excellent feedback from Rick Jelliffe. See below.  /Roger

From: Rick Jelliffe [mailto:rjelliffe@a...]
Sent: Friday, December 30, 2011 9:48 AM
To: Costello, Roger L.
Subject: Re: Engineering versus Science, Anecdote versus Evidence ... [Was: Designing an experiment to gather evidence on approaches to designing web services]

I hope not. Scientific method and engineering method are quite different: one asks what the nature of some reality is, the other asks how we can create something with a certain quality.

The distinction is not not engineering-as-art and engineering-as-science but engineering-using-magic and engineering-using-evidence --where the evidence quality ranges from poor (anecdotal) to excellent (replicated auditable double blind, etc) and the egineering needs to work with all kinds of evidence qualities.

The old saw is that "every schema represents a theory about the document", and every schema comes from some set of concerns. For example, if you were taking a software engineering quality viewpoint for -re-designing a schema, you would identify and weigh the various quality factors for the schema, and design and test it to address the issues of highest concern.  For example, if software maintainability were the main factor, you would look to see where programmers were making mistakes when they write scripts to handle the documents, and arrange the markup so that the element and attribute names and values made it more explicit to the programmer: this would be a secondary concern to coming up with some academically pure but operationally quite bogus model.

For example, the "highly generic schema" approach does that: rather than modeling the data, it tries to model the programmer, in a sense. 

Cheers
Rick

[1] http://broadcast.oreilly.com/2010/07/highly-generic-schemas.html