Created attachment 197123 [details] A diagram with with connectors "merging" into each other before reaching their target Motivation: Typical drawing shapes, not decked with manually-added glue points, can have connectors enter them / connect to them in one of four directions of their bounding-box. This is not so bad for diagrams in which the entrance direction has no semantic significance; or when the in-degree of shapes is at most 1, or when it makes some sense for incoming connectors to "merge" (e.g. a symmetry around the target shape, with two shapes' connectors merging before connection). But - this is really not the general case; and in diagrams with multiple connectors - this looks awkward and indicates a false relation between the shapes being connected to the target shape, as though their connection is through some sort of joining between the connection paths - like river estuaries merging - when the user generally does not intend to convey this sense. It also helps the general sense of a mess of connectors, where it's difficult to associate a line segment back to a connector with its two endpoints, making the diagram more difficult to follow. Suggestion: We should have a mechanism for automatically distinguishing between incoming connectors to the same glue-point or into a shape from the same side. Specifically, the connectors could be made to actually connect at short intervals next to each other, e.g. instead of two connectors ending like this: +---------------+ | | | | +-------*-------+ ^ | they will be rendered like so: +---------------+ | | | | +-------*-------+ ^ ^ | | despite the user having had both of them connect to the center, or just the shape generally from the bottom. Of course this would be an opt-in mechanism, either a feature of the shape or of the connector. One could even imagine assigning connectors to sets such that those from the same set are allowed to merge, and from different sets - made distinct; although that's not absolutely necessary. Naturally this is just as relevant for Draw.
What is the difference to bug 163274? I like the proposal here: Distribute the effective position around a gluepoint depending on the number of connectors. Wonder if this is supported by document standards - likely not. Means we have to automatically add and move gluepoints.
*** Bug 163274 has been marked as a duplicate of this bug. ***