Extended Matrix has several orthogonal grouping registers that all want to “contain” the same stratigraphic units: activity (intention — e.g. “construction of the villa’s external portico”), excavation / location (space — e.g. “saggio 1” / “saggio 2”, the LocationNodeGroup of DP-65 / DP-43), time (EpochNode swimlanes), and paradata (DP-60 per-US / DP-19 swimlane). These registers are genuinely independent: the remains of a single activity are routinely dispersed across several excavation areas, so an activity cluster and an excavation cluster intersect.
yEd — the current EM authoring surface — cannot draw overlapping group folders: a node lives in at most one folder. So the registers cannot all be rendered as folders at once; forcing two (e.g. activity folders + excavation folders) produces a collision. For this reason EM core, while bound to yEd, keeps a single folder register (activity) and represents the others as nodes + edges (is_in_location, is_in_activity, epoch edges) — never dropping the data, only the competing folder rendering. (This is why the shared exporter should serialize LocationNodeGroup as node + is_in_location, not as a second folder layer.)
The future EM graph editor removes the constraint: the user picks which register is the active visual clustering (excavation/location ↔ activity ↔ time), and the editor renders that register as folders / containers while the others stay as edges — switchable on demand, with a layout recompute. The graph data is unchanged across switches (all memberships always present as nodes + edges); only the view changes. This decouples the EM data model (multi-register, m:n) from the single-register limitation of any one drawing surface.
Originating discussion: collision between excavation (saggio) clustering and activity clustering when one activity’s remains span multiple saggi.
1.7
Needed