I once thought this behavior - having multiple line shapes when there was inflection - would let me actually accomplish something useful. My intent had been to create ER models in Barker's Notation (a la David Hay et al). This notation requires connectors to be drawn such that source and destination 'halves' may use different line patterns (solid, dashed). I figured I'd just force a line inflection (even if the line ended up 'looking' straight, and let a shape script take care of each segment as required.]
[Yes, I hear you, using a side effect of a design quirk is bad. But finally we had a rigorous treatment of metamodels in the context of Zachman, and that's just about the only thing we really have to come up with, and here was a major stumbling block for using EA. It seemed that this unintended behavior might actually let me get things moving again.]
Anyway, not only does EA create two 'shapes' and draw them without 'telling' the designer, it seems that this behavior is completely 'canned' from the designer's point of view. We cannot get at the 'inner' shape engine and provide our scripts for line style. So both (or all) of the shapes used in the line are the same.
[So EA saved us from the bad way of going about things after all. The result is that we simply cannot do the job with EA, at least not in the way that the only (modern, rigorous) published treatment out there uses. Not good; not good at all.]
So, ideally we want the 'bug' resolved, but not by removing the extra segments. What we really need is the ability to define the portions of the line differently (or at least provide the shape script the ability to see the start, end, and any middle portions as different line shapes, and to provide scripts for each of these. We also need some means of referencing the line segments created by inflection points. This could be indirectly through the shape script, or ideally through both the script and an extension to the API. [This latter case could be handled without changing the schema. The segments could be under the control of EA, and not treated as a 'first line' collection.]
Another great benefit would be that finally we would have a predictable means for breaking a line between two widely separated elements and 'removing' the middle segment from the display. When using automation or profiles this would be far cleaner than the current situation.
I need to stop dreaming now. It's time to wake up...
