The Extended Matrix has been using the word anastylosis in two different senses without acknowledging the difference. In one sense — the strict archaeological-conservation sense — an anastylosis is a present-day reassembly of physically surviving fragments: a column drum is repositioned on its base, a frieze block is replaced on its course, a statue’s lost arm is propped against its torso, a Greek temple is hand-rebuilt out of its own scattered stones. The fragments are real, they exist today, and what’s being modelled is the modern act of putting them back together. In a second sense — closer to a virtual-reconstruction workflow — the word has crept in to describe the past-state reconstruction itself: hypothetical pieces (VSF, USV*) are added alongside the surviving fragments to complete the building as it stood in antiquity. The modeller is no longer working in the present; they’re rebuilding a past state. The two activities share a verb but not a temporal anchor, and a single group-node class glossing over the distinction has masked a real semantic boundary in the language.
DP-68 separates them. AnastylosisNodeGroup is the present-day case, the group that aggregates real fragments (SF + RSF) into a current configuration. RestorationNodeGroup is the past-state case, the group that aggregates virtual elements (VSF + USV*) together with whichever real elements participate in the past state. The discrimination rule is sharp: the presence of a single VSF or USV member flips the temporal anchor of the group from present to past*. Adding a VSF to an AnastylosisNodeGroup is therefore a validation error — not because the modeller’s intention is wrong, but because the group’s identity changed under their feet and the language should make that explicit.
Anastylosis as a study. The third refinement that surfaced in conversation is the most interesting of the three. A present-day fragment assembly — even one that’s never going to leave the modeller’s desktop, never going to be physically built, never going to be a museum exhibit — can be the act of generating evidence for a past-state reconstruction. The modeller assembles the fragments today, learns which pieces fit, identifies the missing ones by shape and join, deduces the column’s height from the diameter of its lowest preserved drum — and that knowledge feeds the RestorationNodeGroup. DP-68 formalises this dual role: an AnastylosisNodeGroup whose study_role is set to true becomes a complex source in the paradata chain of a downstream RestorationNodeGroup, plugged in via the is_premise_of edge from DP-66 (the same edge that lets a PropertyNode of one stratigraphic unit serve as a premise for a Combiner producing a property of another). The Anastylosis Study isn’t a separate class — it’s an AnastylosisNodeGroup playing a specific epistemic role, expressed through its outgoing edges rather than through a class change.
New-from-virtual: the Tommaso & Sara case. The fourth pattern emerged from a real fieldwork situation: a present-day physical construction that completes a past Virtual SU. The modeller documents a brand-new course of bricks they’re laying today, but the reason they’re laying it is to anastylosise — physically, in place — what the RestorationNodeGroup says was once there. The new bricks are not virtual: they exist, they have an excavation date in 2026, they would show up in any future stratigraphic survey of the wall. But their epistemic motivation lives in the past — they’re there because the past USV said they should be. DP-68 captures this with no new class: the present-day US is a regular US, but it carries a was_derived_from edge (the property inheritance edge from DP-66) pointing back to the source USV in the RestorationNodeGroup. The graph then expresses cleanly: ‘this modern wall is the physical completion of that hypothesised ancient wall’, queryable end-to-end, without conflating the two temporal anchors. The same edge type that lets a property be inherited across containment in DP-66 also lets a present-day stratigraphic unit declare its derivation from a past one — same pattern, applied at a different granularity.
Why this isn’t just cosmetic. A modeller working a Roman temple has, today, several distinct things they might be doing: (a) excavating fragments and recording their find-spots (regular US authoring); (b) reassembling those fragments on the museum floor or in the viewport, learning from the fit (AnastylosisNodeGroup); (c) hypothesising the original composition (RestorationNodeGroup); (d) physically building a partial modern anastylosis in place to consolidate the surviving structure (new-from-virtual, regular US linked via was_derived_from to a past USV). Conflating (b) and (c) under one label, as the EM 1.5 vocabulary effectively does, makes (d) inexpressible — the present-day modern build can’t claim its motivation in the past virtual without polluting either group. The cleanest formalism gives (b) and (c) their own group classes, lets (a) and (d) live as regular stratigraphic units distinguished only by the presence (or absence) of a was_derived_from edge, and connects (b) → (c) when needed through is_premise_of. Four behaviourally distinct workflows, three syntactic constructs (two new group classes + the existing US class with two existing edges from DP-66), no semantic overlap.
CIDOC alignment falls out cleanly. AnastylosisNodeGroup is a present-day activity that physically rearranges material: maps to crmarchaeo:A5_Stratigraphic_Modification. RestorationNodeGroup is a hypothesised past activity that constructed something: maps to crmarchaeo:A4_Stratigraphic_Genesis, with the temporal anchor in the past. Anastylosis-as-Study layers crminf:I5_Inference_Making on top: the present-day assembly is one of the premises (J5_used_as_premise) of the inference that produced the past-state reconstruction. The new-from-virtual case is a present-day A4_Stratigraphic_Genesis (a real construction event today) whose causal motivation is captured via prov:wasDerivedFrom (the RDF mapping of the was_derived_from edge from DP-66) pointing back to the past A4 it’s completing. All four patterns sit on top of CRMarchaeo / CRMinf / PROV-O without forcing the EM extensions to invent new top-level predicates.
Validation rule, expressed in plain terms. The single rule that captures the whole discrimination: the temporal anchor of a stratigraphic group node is determined by the membership of any USV or VSF — if the group contains even one such member, it’s a Restoration; otherwise it’s an Anastylosis (whether playing study role or not).* The EM Tools UI surfaces this as a one-click ‘convert to Restoration’ when a modeller drops a VSF into an Anastylosis group, so the validation is helpful rather than punitive. The reverse direction (dropping a USV-typed member out of a Restoration, leaving only SF/RSF behind) offers a symmetric ‘convert to Anastylosis’ suggestion.
Why this lands in 1.7 rather than 1.6. The work is a language extension (two new group classes), a UI refactor (Visual Manager toggles, Add-Group dialog updates, validation messages), and a migration path for legacy AnastylosisManager projects. None of these are trivial. The 1.6 cycle is already absorbing the canonical-edges series (DP-62), the multilingual labels (DP-63), the property inheritance pattern (DP-66), and the USNV class (DP-67) — DP-68 stacks on top of DP-66’s was_derived_from and is_premise_of edges and on top of DP-67’s experience with introducing a new stratigraphic-family class, so the 1.7 cycle is the natural home for it. The conceptual work captured here is the design specification; the implementation work will follow once 1.6 closes.
Open decisions for the review call. Five points to thread at the review meeting: (1) the study_role enum on AnastylosisNodeGroup (proposed: mere_assembly / study / display); (2) Activity context for new-from-virtual physical US (proposed: separate ‘restoration-in-progress’ Activity, not auto-joined to the source USV’s Activity); (3) one-click ‘convert to Restoration / convert to Anastylosis’ in the validation UX (proposed: yes); (4) yEd palette colour calibration for the two new group identities (deferred to DP-30 palette work); (5) cross-graph anastylosis (a fragment in graph A participating in an anastylosis group in graph B, e.g. museum-floor reassembly pulling fragments from multiple sites) — out of MVP scope, would need a multigraph-aware membership edge building on DP-10. The TempluMare and Basilica Julia projects are the natural early stress tests for the Anastylosis Study workflow; Tommaso & Sara’s Turkish fieldwork is the canonical new-from-virtual case the language should support cleanly on first use.
1.7
Needed — Tommaso & Sara (Turkey) for the new-from-virtual bridge; TempluMare or Basilica Julia for the Anastylosis Study workflow stress-test
Designed in conversation 2026-06-09 from a brainstorm where the maintainer identified that the existing informal ‘anastylosis’ language confused two structurally distinct cases (a present-day fragment assembly vs a past-state reconstruction), and surfaced a third pattern (present-day new physical construction motivated by past Virtual SU) that doesn't fit either of the first two. The discrimination point is sharp: **VSF inside the group flips the temporal anchor from present to past**, so a single class can't cover both. The user's working example: ‘imagine you have SF, VSF, USVs, USVn, USVc, all together they are blocks creating an anastylosis group. Now this anastylosis group is something I can do today using the blocks, as first instance. It's an activity to build with old blocks, so I give them a new position and therefore I have to use Reused Special Finds. The VSF I can use or not, actually I shouldn't use them because the moment I start using virtual special finds I'm also working in the past. And they become a restoration node group.’ The Anastylosis Study refinement followed: ‘L'anastilosi study is a simulation with the fragments that happens today and can be used as a complex source for the ancient reconstruction. I'll understand it by working on TempluMare or Basilica.’ The new-from-virtual bridge surfaced last: ‘So I do a simulation today, as if I wanted to reconstruct, and instead of using virtual structural or non-structural stratigraphic units and virtual special finds, I create new stratigraphic units, which are non-virtual but are physically virtual. This is interesting, in the sense that I'm creating them to complete something that existed in the past. I could simply develop them from the virtual one. This is perhaps the best thing, so I link the past virtual stratigraphic unit, object of study, to a new construction, so new stratigraphy that I'm going to document. For example in the case of Tommaso and Sara who are doing it in Turkey.’ Open decisions to thread at review call: (1) the `study_role` toggle on AnastylosisNodeGroup — should it be a boolean flag, an enum (`mere_assembly` / `study` / `display`), or expressed only through the presence of `is_premise_of` edges to a downstream Restoration? Proposal: enum, because three distinct curatorial intents already surfaced. (2) When the new-from-virtual physical US is built, should it AUTOMATICALLY join the same Activity as the original USV (inheriting temporal context), or should the modeller place it in a separate ‘restoration-in-progress’ Activity? Proposal: separate Activity, with the `was_derived_from` edge as the link — preserves the chronological distinction between ‘the past wall’ and ‘the modern wall built to complete the past one’. (3) Whether to formalise the AnastylosisNodeGroup → RestorationNodeGroup ‘upgrade’ as a one-click operation in EM Tools (when a modeller drops a VSF into an Anastylosis group, the validation message offers ‘convert to Restoration’ as a one-click fix). Proposal: yes, makes the validation rule helpful rather than punitive. (4) The yEd palette colour proposals (brown-orange for Anastylosis, gold for Restoration) need calibration with the existing EM palette — flag for the palette work in DP-30. (5) Cross-graph anastylosis: a fragment SF lives in graph A (its excavation context), but participates in an AnastylosisNodeGroup defined in graph B (a museum-floor reassembly that pulls fragments from multiple sites). Out of scope for the MVP; a multigraph-aware membership edge would need a separate design call building on DP-10 (Multigraph Project). Cross-refs: DP-43 (Group Nodes — the parent class of both new group types), DP-26 (Spolia / RSF — RSF is a primary member of AnastylosisNodeGroup), DP-46 (Proxy Box Creator — produces the SF/RSF proxies the AnastylosisNodeGroup aggregates), DP-50 (Surface Areale — alternative proxy strategy for restoration walls), DP-60 (Paradata NodeGroup — sibling group concept), DP-66 (Property inheritance + premise reuse — the `was_derived_from` edge for new-from-virtual + `is_premise_of` for Anastylosis Study), DP-67 (USNV — adjacent stratigraphic class formalisation), DP-10 (Multigraph — eventual home of cross-graph anastylosis if pursued), DP-30 (Reconstructive Elements — palette calibration), DP-23 (Music Reconstruction) and DP-24 (Dance Reconstruction) as conceptual cousins (present-day instantiation of past phenomena, possibly inheriting a similar past/present scoping pattern).