Proofs of concept
A map of a legacy software system
An in-progress study of interconnected repositories from an older project, using their structure and history to inform a new commercial architecture.
Status: proof of concept in progress. The project, repositories, code, owners, and prospective commercial system remain undisclosed.
The starting material
A set of repositories from an older software project: source trees, package manifests, configuration, documentation, commit history, and observable dependencies. The goal is not to revive the old product unchanged. It is to understand how the pieces interacted and what that history can teach a new architecture.
The graph question
Can repository evidence become a system map that distinguishes implemented coupling, intended boundaries, historical accidents, and reusable domain knowledge?
Possible entities include repositories, services, packages, domains, data stores, interfaces, deployments, teams, decisions, dependencies, events, and capabilities. Relationships might include “depends on,” “publishes to,” “consumes,” “implements,” “owns,” “replaced,” “shares schema with,” and “constrained by.”
What the study tests
- parsing code and configuration across several repositories;
- identifying runtime, build-time, data, and organisational dependencies;
- separating explicit documentation from inferences drawn from code;
- connecting architectural decisions to later consequences;
- locating duplicated concepts and hidden coupling;
- preserving private source boundaries while publishing only reviewed abstractions;
- using the map to compare possible boundaries for a new commercial system.
Potential working experience
A private technical knowledge graph where an architect can move from a capability to the repositories, services, schemas, and decisions that implemented it. Public output, if any, would describe the method and reviewed architectural lessons—not expose proprietary code or confidential details.
Learning goal
This study tests repository-scale ingestion and architecture recovery. It asks whether a governed graph can turn legacy evidence into a more inspectable foundation for new design decisions without mistaking generated interpretation for fact.