20th March, 2003 
Reality bites: theories into practice

I was thinking about theoretical ideas that failed when put into practice and wondered why they didn't work. It seems sometimes the problem domain expands to cover impractical situations - those situations that never occur in real-life. Sometimes the problem domain is simplified to fit the solution. And sometimes the model doesn't account for practical implementation.  
  
Electronic Data Interchange (EDI) seemed a logical idea - to provide a universal language for companies to exchange transactional information such as purchases and invoices. We assumed with some naivety that the language would be simple and straightforward.

However the information content was broadly variable so the vocabulary for communication was large, even though practical communication between individual trading partners used a limited vocabulary. This situation strayed further from the ideal with the push for dialects, where different industries imposed variations of the language structure. For suppliers to multiple industries, the outcome was modified information exchange on a case by case basis. In retrospect, it would have been easier to allow laissez faire business forces to shape the development of the exchange of trading information - at least the simplest practical language would have evolved.

The business premise for EDI was that the simplest and hence most cost-effective approach for companies to communicate would be to speak a third, common language that all understood. However, the starting condition was that no one spoke that language. Suppose that two people wanted to communicate with each other, one an English speaker and the other a Japanese speaker. Very rarely would there be a practical reason for speakers to choose to learn Swedish in order to communicate.

It was argued that EDI would provide economic benefit for all participants. It would reduce manual entry of information saving time and effort, and reducing entry mistakes. The generalisation in this assumption failed to account for smaller trading partners. For these partners whose trading volumes are smaller, the cost of installing, integrating and maintaining a complex EDI translator is not offset by the savings of automatic entry.

The OSI 7-layer network communications model is an example of imposing logical organisation without accounting for the costs of implementation. The model provides an ideal separation of functions, easing the understanding of the individual tasks of the layers. In implementation, the 7-layer model introduces complexity and inefficiency. The model for its aesthetic appeal, does not translate to a practical solution because it hides important considerations in implementation.

Of course, it is easy to criticise in hindsight. The real point is that we can deceive ourselves if we do not question assumptions or test the veracity of a model, be it business or technical. It is always worthwhile to test against real world situations or to compare a similar real world situation with the operation of your model. It can give you some valuable insights into whether the proposal will work and whether your assumptions bear out in the field.