Re: Revised diagram following today's discussions

I think I'm slowly starting to understand this diagram. I am still not
sure about FlowGraph and Flow though.

Anyway, I tend to think of pipeline analysis in phases that are
relatively discrete and separable. I don't know if this will resonate
with anyone else, but here goes anyway.

1. There exists some concrete syntax for a pipeline document. That
document contains steps and bindings for inputs, outputs, and
parameters.

2. From that document, we can build a set of component declarations.
Some of these come from p:declare-component elements, others are
implicit (what Alex called component types, I'm pretty sure) in the
use of "step container" steps.

3. The component types have declared inputs, outputs, and parameters
which means each component type has a name and signature.

4. The input, output (and perhaps parameter) bindings that a step
specifies are constrained by the position of the step in the document.
(Is it a step in a step container, for example).

5. Steps are instantiated into one or more components. These
components are the things that appear in the flow graph, they have
pipes coming into and out of them.

6. Every step that instantiates a component must have bindings that
are consistent with the signature of that component.

Dunno if that helps anyone but me. (Or points to where I'm clearly
engaged in wrong-headed thinking.)

                                        Be seeing you,
                                          norm

-- 
Norman Walsh
XML Standards Architect
Sun Microsystems, Inc.