Structure and Dimensions

Brad acknowledged the anti-thesis and it is fine. Between what and what is a matter of terminology, which is important, but will be dealt separately (something about it in refexions on agility— much later than this discussion).

However, the second of Brad's sentences (in his first paragraph), about the dimensions of the composition, contradicts the necessity of genericity of the structure (this had been superficially treated while speaking of configurations items).

CM may compose and decompose at will (width first), but SCM has to be (depth first) both more humble, and more ambitious, and to focus on a structure which would leverage its efforts, and thus be adapted to its tools. Such a structure must thus be generic, which excludes all of the "physical, functional, historical, environmental, and managerial" dimensions. Going on this route would be dispersing one's efforts.

One (not necessarily the only one, but the only supported by tools such as ClearCase audit records) single dimension which offers such a genericity, is this of dependencies (again understanding by this only what can be audited in an automatic and objective way). This one builds obviously upon the very primitive, but very powerful semantics of sameness and difference.

It is symptomatic that Brad in the following described the intricate problems raising from the contradictions between the various specific dimensions (and forces). Their resolution is indeed SCM goal, but SCM can only achieve this goal by taking a neutral and objective position, and refusing to play the political games of trading off between incommensurable concerns. There shall of course remain in the end a 'political' decision, but this is out of the realm of SCM: this one is only there to produce the facts (in context, under the constraints of the situation: so nothing universal and absolute, but something clear and common, inescapable, on which to agree, a basis for the discussion).

On the issue of decomposition and recomposition, Brad is right in that both CM and SCM share the same traditional, rational, and modern (in the historical sense of the word) approach known as divide and conquer. Where they depart from each other is about avoiding the pitfall of reductionism (also known, per Karl Popper, as historicism), i.e. confusing one's model with reality (experience). When attempting recomposition, SCM is thus focusing on reproducing an audited process, maybe with incremental changes, but in a humble and descriptive (non intrusive) way, attentive to the results. CM would on the contrary tend to apply cook book recipes, and forbid any other method (in a prescriptive way).

-- MarcGirod - 08 Feb 2007