Overview
    by Stan Hendryx (Chair of the BRWG)

In this working group^[1] we are trying to write down requirements for Business Rules in models that could be the basis for standards for Business Rules at the business perspective, including how to get to other perspectives.  I have heard people who appear worried about whether we are going to be "pure" to the Zachman Framework -- in other words, true to your notions of "architecture." 

Can you comment on the relationship of your architecture framework to other architecture frameworks.  How do you rationalize your architecture to the others?  (I've mentioned a few.)  It seems like we need to be able to align and rationalize 'frameworks' if we are going to standardize.

Response
    by John Zachman

When you think about the enterprise it is complex.  Yet people want to try to understand it in a simplistic way.  We want ways to "simplify" without be "simplistic."  I think my " architecture framework" does this. It has these characteristics -- it ...

• is stable ~ it has been around for a LONG time
  
  
• uses a logic that is fixed
  
  
• has a sense of completion (completeness)
  • without this, the next change/silver bullet will kill you

This gives us a good structure to work within.

First, it gives us six primitives for the columns (interrogatives) that form a complete set for describing any thing.  If you add columns, you produce a "denormalized" structure -- things beyond these primitives.

Then, there is the other dimension, which gives us three solid perspectives -- conceptual, logical, physical -- plus two "bounding" perspectives -- scope and out-of-context.  Many people confuse this with "level of detail" and that is wrong!  These are different models ... different perspectives.  Transformations need to occur between perspectives.

Furthermore, it is not a "decomposition."  Decomposition occurs "within a cell."  Every column is an "independent variable"; every cell (therefore) is an independent variable.

I am confident about the classification scheme itself.

Follow-on Exchange (Q&A)
    Questions fromthe audience -- Answers by John Zachman

Q:  Could you take another classification scheme and make some mapping to it?

A:  Assuming that it is primitive and elemental, yes, you should be able to map any description to it.

Q:  I have been asked "how many models" does it take to model a row?  To be complete, you would need "6 models" with the relationships between them.  People seem to think in terms of "diagrams" -- could you have a "diagram" that addresses more than one model?

A:  Every cell has been described so that it (its model) is unique/discrete.  But for implementation purposes, you need composite (non-primitive) models.  For architectural purposes, you want the primitive models.  These are the models you must have for:

• alignment
• reuse
• reduced time to market

It is important to note that an enterprise-wide composite model becomes so complicated you can never get your mind around it; you need to factor things out -- get the primitive structures defined.  We need to engineer the primitive constructs so that we have them for reuse (etc.).

You can think of this like the periodic table:  until you know the elements, you are only guessing at the compounds.  Repeatability comes at the elemental level.

Q:  For the items that populate each of the cells (or even for the composites), is there a requirement that they be "graphic"?  In other words, is there a dictum of the Zachman Framework that says that any model MUST be graphical?

A:  Not necessarily -- but by typical definition, we wouldn't consider a "list" to be a model.

<sh> comment:  The OMG says that it must have an underlying structure to make it a "model."

Q:  Can something expressed ONLY in a non-graphical form still be a model?

A:  If you had a complete (holistic) set of graphic models across an entire row you might not need text.  But that doesn't mean that the model(s) could not be expressed in text.

Q:  Are the 'rows' the conceptual/logical/physical constructs - or are they a description of the viewer (perspective)?

A:  The perspectives characterize the models of the row; they do not describe "who owns" the model. For example, I had a business person (CEO) who felt he "owned" the information system data models and that this meant the data models were "row 2 models." An information system data model is an artifact of row-3/column-1, regardless of who "owns" it.

Q:  I have found it helpful to characterize it as "whose purpose is being served"; i.e., it isn't saying that this model "belongs to" the Owner but rather that it serves the purpose/perspective of the Owner.

A:  Yes, and also the model at row-3 needs to support THE INTENT of its corresponding model at row-2 (etc.).

Summing up
    by John Zachman

Long ago, I got connected with the IRDS effort (where I met Sowa, a philosopher).  That work dealt with four levels of 'meta.'  I took the logic structure and applied it to the enterprise -- but it is the same analytical structure at any 'meta' level, just with different analytical targets/products.

However, the world wasn't ready to deal with things like this at that time.  Maybe it is now....

References

[1]  BRWG (Business Rules Working Group). This OMG activity has since been defined as a SIG -- the Business Rules SIG.  Its work is reported periodically in the Business Rules Journal. For example, see www.BRCommunity.com/a2003/b139.html. 

# # #