ASSESSMENT · EVIDENCE · IMPLEMENTATION
From scattered ruins
to connected readings.
The original prompt identifies a real interpretive problem and proposes a useful response. An interactive GIS can connect dispersed areas, explain their different histories, and anchor photographs and discussion to specific places. It becomes feasible when geographic records, published phase evidence, observed fabric, reconstructions and visitor contributions are represented as distinct records.
What the original prompt gets right
Spatial distribution and temporal layering need to be studied together. Carthage is not one intact exposed city. UNESCO’s component inventory and integrity account establish the variety of surviving areas and urban pressure. The atlas therefore navigates among sites rather than presenting a single, supposedly complete ancient city plan.
“Topological uniqueness” needs a more precise definition. Here it means relations such as overlies, reuses, beside, within a district and photographed from. Terrain elevation, slope, intervisibility and drainage are topographic analyses and need independently controlled elevation data. This release calculates distances between cartographic anchors and presents a documented ordinal succession; it supplies no invented elevations or viewshed results.
Four necessary corrections
| Original idea | Logical constraint | Concrete redesign |
|---|---|---|
| Map the dispersed ancient city. | A modern mapped park boundary is not an ancient building footprint. A camera location is not a subject location. | Store provenance and geometry type. Use current OSM context with source element IDs. Keep one approximate street anchor visibly distinct. Export anchors and mapped extents separately. |
| Use images / 3D to show stratigraphy. | A photograph records an exposed view, not a complete sequence of deposits. A phase chronology is not a trench-specific Harris matrix. One picture cannot supply measured 3D geometry. | Pair photographs with published phase claims and a source locator. Use region annotations and comparative views. Label museum models as modern interpretations; show no fabricated wall heights or buried layers. |
| Visitors and experts enrich features. | Self-declared expertise does not certify a reading. Contradictions and source requirements need durable records. | Persist image targets, normalized regions, names, evidence categories, citations, replies, review notes and version history. Contributions begin pending. Curator decisions do not silently rewrite seed claims. |
| Build the synergy into HTML. | Static HTML alone cannot provide durable cross-device collaboration. | Use HTML for the interface and a local service for shared storage. Public deployment later needs accounts, authorization, moderation, backups and agreed image rights. |
Source coverage
The cited bibliography supports targeted readings, not a claim of complete coverage. No source publication is republished.
Targeted published readings
The first prototype inspected five publications at passages needed for seed claims. The history extension expands this to eight targeted publications and a bibliographic correction of a ninth. Other distinct files were inventoried and text-extracted; they have not been fully read or treated as claim authorities. The in-app library exposes this difference.
| Publication and locus | Finding used | Effect on design |
|---|---|---|
| H. Hurst, Understanding Carthage as a Roman Port (2010) ; printed pp. 53, 55–56, 65; Fig. 5 visually inspected. | The naval harbour and Roman island complex require distinct readings. Harbour function and coastline development remain interpretive questions. | Keep separate phase cards. Pair the supplied museum-model photographs with Hurst’s discussion. Do not flatten commercial and state-linked uses into one uncontested explanation. |
| B. Kaufman et al., Ferrous metallurgy… (2016) ; JAS 71 pp. 33–35; Fig. 1 visually inspected. | Metallurgical claims depend on analysed residues from excavated contexts. | Do not identify an industrial installation by visual resemblance alone. Retain sample, trench and activity as future distinct entities. |
| R. F. Docter, Carthage and its Hinterland (2009) ; printed pp. 179–181, especially table and note 12. | A conventional Punic period table is preliminary and accompanied by a radiocarbon caution. | Retain named chronological schemes. Atlas bands are approximate reading aids, not exact dates for every feature. |
| D. Hoyos, Carthage: A Biography (2021) ; printed pp. 17–18, 94–96. | Byrsa was substantially reshaped; the Tophet raises different questions about remains, literary reports and cause of death. | Use qualified synthesis for orientation. Link conflicting scholarly positions; do not make a photograph settle an osteological debate. |
| Kelsey Museum, Carthage Then and Now (1979) . | The Save Carthage campaign and earlier photographs record a changing history of excavation and presentation. | Add investigation and restoration history to a future phase model. Treat the brochure as a dated source, with its own research context. |
The files labelled Lézine (1968), Mills (2013), and “1964 ben khader…” are promising further research leads. Their detailed plans, fabric arguments and bibliographic identity were not fully audited here. In particular, the last filename’s date remains unresolved. No copyrighted publication plates have been copied into the public asset tree; originals can be consulted only through the loopback service on this computer.
Accessible precedents and adopted patterns
| Resource inspected online | Pattern adopted | Boundary |
|---|---|---|
| DAI / INP: Quartier Didon | A documented succession, uncertain functions and an explicit distinction between reconstructed and original fabric. | The Roman hall function is unknown; the rotunda’s cult use is probable. The street anchor is approximate. |
| AMVPPC monument pages and media | Area-specific interpretation and links to photographs, videos and virtual tours. | The Tophet overview is sparse. The bath abandonment date is unknown. These pages do not supply a universal survey model. |
| Recogito | Point / rectangle selection, source-place links, collaborative readings and export. | The implementation is an independent local service, not a connected Recogito workspace. |
| UCLA Digital Roman Forum | Reconstruction accompanied by an explicit represented date and evidence documentation. | The project description was verified; availability of all legacy model interfaces was not assumed. |
| British Library / LaNC: Peripleo | Map and collection browsing reinforce each other. | A collection explorer does not validate the archaeological interpretation of an individual object. |
| iDAI: Karthago Volume 1 | Chapter-level links for plans, building techniques, Roman phases and presentation. | Landing-page and chapter-list inspection only. The 2026 digital publication date must not be confused with excavation dates. |
London Charter principles 2–6 guide source evaluation, documentation, preservation and access. W3C Web Annotation supplies the export pattern: a textual body targets a specific image region. The export uses pixel xywh selectors tied to an immutable display-copy hash. The IIIF selector model informs future interoperability; an IIIF image server is not implemented.
Feasibility by stage
| Stage | Feasible outcome | Inputs still needed |
|---|---|---|
| 1 · Current release | Linked map, sourced phase reader, supplied image collection, spatial anchor comparison, shared local annotations, replies and review. | Continued expert identification of images and individual building features. |
| 2 · Survey and stratigraphic GIS | Georeferenced historical plans, trench/feature records, phase-specific geometry, explicit stratigraphic relations and elevation comparison. | Plan reuse permissions, control points, CRS, residuals, trench/context identifiers, measured elevations and published phase crosswalks. An ordinal reader is not sufficient input. |
| 3 · 3D study | Measured present-condition models; separate source-backed reconstruction variants with selectable evidence confidence. | Overlapping calibrated images or scans, scale/control targets, camera and acquisition records, coverage/accuracy assessment, model rights and per-element reconstruction paradata. Single photographs and photos of models cannot supply this. |
| 4 · Public collaboration | Cross-device contributions and expert curation under a published governance model. | Hosting, verified user/curator accounts, authorization, moderation responsibility, backups, licensing decisions and an agreed data-release scope. The current loopback server must not be exposed directly. |
Audit findings and remaining uncertainty
- UNESCO’s English text conflicts internally between 146 BC and 146 AD. The seed chronology retains 146 BCE and records the erroneous wording rather than silently relying on it.
- The Quartier Didon research record is linked to the Rue Ibn Chabbat street, not the unrelated Villa Didon hotel. No precise park footprint has been asserted.
- OSM contains duplicate representations and suspect shared Wikidata identifiers around the Odeon / Roman Villas. External IDs are not automatically treated as archaeological equivalences.
- OSM data are downloaded modern context, with element provenance and ODbL attribution. They supply no official UNESCO boundary and no phase-specific ancient plan.
- The three orientation-only entries—La Malga, the amphitheatre and Damous el Karita—remain without detailed phase claims. Their absence from dated filters is explained as incomplete evidence entry.
- Punic and Roman museum models can be compared as hypotheses. Their elevations are not measured reconstructions generated from the field photographs.
- All annotation names and expertise are self-declared. Curator review status is procedural metadata, independent of the claim’s evidence category.
- Photograph reuse beyond local study is unresolved. Rights and source checksums remain in the data, and publication originals are outside the static asset tree.
- Recent online information was checked on 1 October 2026. Live fees and opening hours are deliberately excluded because the application is a research reader rather than a current visitor timetable.
Run, inspect and continue
cd carthage-gis python3 server.py --port 8765 # Open http://127.0.0.1:8765 # Validate source links, geometry, image dimensions and hashes: python3 scripts/validate_data.py # Exercise the actual API with isolated test storage: python3 scripts/test_service.py # Rebuild supplied-image derivatives and seed data from the recorded local inventory: python3 scripts/build_data.py
The local service requires Python 3.9 or later and Pillow. Leaflet 1.9.4 is bundled with its license. The application is usable without external map tiles after the local server starts; links to external scholarship still require connectivity. private/ contains the bounded inventory, extracted research text, original-reference path map, contribution database and curator token. It is excluded from static serving. The source library endpoints intentionally permit consultation of selected source edition through the loopback service.
For a portable public package, remove private/, decide which photographs may be distributed, review identifiers and phase ranges, and deploy a separate authorized service. GeoJSON and Web Annotation exports preserve provenance; the custom review properties use a proposed project namespace and do not imply an existing public vocabulary service.