Date: 25 Aug 96 23:20:59 EDT
From: Martin Fowler <100031.3311@CompuServe.COM>
To: Thierry Thelliez <>
Cc: Jeff Sutherland <>
Subject: Re: Patterns and Business Objects


> I am interested to know more about why such efforts are
>'doomed' and how Business Objects can still help in this domain.

Well there's no theoretical reason why they are doomed. The reason is just that in my observation I've seen many failures and no real successes.

> Your idea of patterns is seducing but I was wondering how
>you can practically use it. I could imagine how to use and reuse
>software components (implementing Business Objects) in the
>development of an IS. I could also imagine that during the Analysis
>phase of a reengineering effort, we could recognize some patterns
>across different domains/applications. I have more difficulties
>understanding how Analysis Patterns could help in reconciling opposite
>business views.

Assume one group builds an accounting framework which they want to spread around an organization. This framework will typically have many classes and in effect contain many analsysis and design decisions. (Business Objects rarely stand in isolation, in practice most business objects will have many links to others.) A department who wishes to use the framework has to accept the whole thing. So they look at the decisions that are in the framework and if they sufficient points of discomfort they will not use the framework.

With patterns however its not an all or nothing decision, they can use what they like and change what they like. The result is not as good as if they had adopted the framework, but not as bad as ignoring the framework and not using any patterns.

Of course a really good framework will be sufficiently malleable that users can override all the decisions that they don't like (ie provide a suitably abstract account or building). But building such frameworks (and making them comprehensible) is a difficult task. Patterns provide an interim measure which I belive will help us get there.

> Do you have some examples of successful implementations ? >Is your book going to cover that ? When will it be available ?

I must stress this notion of using patterns as informal standardisation is pure conjecture on my part. It sounds like it might work, but its still too early to see any evidence. (Although they way people have adopted and use the GOF patterns suggest it could well happen.) My experience so far is really in using the models to help the design work that I and my clients have done. So the book does not try to address the informal standardisation issue. The book will be out at OOPSLA.