Generic Semantics

SCM is more than just accounting: it can bring in semantics.

These must only be generic (with respect to the specificities of what is being managed). At the bottom, they build upon sameness, and can be leveraged through structure. The most convincing generic structure I can think of is this of (audited) dependency trees.

The structure commonly used, on the other hand (e.g. ClearCase UCM), is this of the original project, typically mapped onto a directory structure. I believe this is specific, both with respect to the history (with is typically only an accident from the point of view of the resources), and to the syntax: this kind of hierachies is only in directory elements, which will lead to discontinuities.

SCM is on opportunity to switch from the traditional production to a maintenance perspective.

-- MarcGirod - 26 Jan 2003

Interesting about the audited dependency trees. Do you mean physical dependencies (e.g. build/compile/link dependencies), or are there more than just that? Your mention of UCM makes me think of UML, whose creators speak of a 4+1view of architecture

Susan Dart describes SCM "functional areas"^? (of which accounting is only one) that look to be even more views of.

Are you refering to all of these at a systemic-level or are you focusing on some particular subset of them?

-- BradAppleton^? - 26 Jan 2003

Lots of conceptual material --I knew Kruchten and was aware of Dart, but had to read Tu & Godfrey. Lots of concept overloading as well, unfortunately. Especially, my use of auditing does quite obviously not fit in Dart's. So I should answer that I meant physical dependencies, and no particular subset of Dart's views.

-- MarcGirod - 28 Jan 2003

One problem I have with both of these classifications, is that the look at SCM from the outside, they are meta-information, and thus themselves not managed.

The meaning they try to reach is global, but itself not grounded: I could think myself as adding more views to the 4+1 model (as do Tu & Godfrey), such as a packaging view between their build-time and the deployment views. Similarily, I'd wonder whether Dart's views are independent, whether they build up dimensions in a mathematical sense, and whether they generate a complete and consistent system.

I think generic semantics escape this critique. Maybe Kruchten's scenarios/contextual view is in fact also an attempt to genericity, more than yet another meta-level.

-- MarcGirod - 30 Jan 2003