Specify the work
Express the intended outcome, satisfaction conditions, delegated authority, and constraints.
INSIDE SPAN
The network evolves.
From moving data
to coordinating work.
SPAN is the semantic execution fabric for autonomous entities.
A work request describes the outcome and the conditions for satisfaction. SPAN coordinates the capabilities needed to achieve it, subject to the authority and policies of the participating domains.
THE ARCHITECTURAL SHIFT
The state belongs
to the work.
SPAN’s architecture centres on a distributed logical state machine for each work request. The coordination state records what has been accepted, what remains to be done, and what may happen next.
Individual capabilities execute locally. The fabric coordinates their contribution to the overall outcome, including changes, approvals, and evidence.
SPAN coordinates meaning, not just data.
HOW THE FABRIC WORKS
One request.
Explicit conditions for success.
Discover and bind capabilities
Resolve suitable advertised capabilities. Check the participants and the proposed work against applicable policy before binding execution.
Execute within each domain
Local execution environments carry out authorised work. Inline controls and runtime integration provide the enforcement points.
Accept evidence and advance
Evaluate returned evidence, update the work state, and determine the next permissible transition. Completion is assessed against the original request and any authorised revisions.
COLLABORATION WITH LOCAL CONTROL
Share capabilities.
Govern access to the data.
Organisations can work together without requiring a common internal data model.
The design separates descriptions of capabilities from the local facts and resources used at execution. Shared semantics describe what can be done; local policy governs how data is accessed and what may be released.
For technical readers: the TBox supports capability descriptions and graph construction; the ABox remains local and is bound under policy at execution.
End-to-end control depends on integrating the relevant execution environment, harness, and sandbox. Active work also needs runtime cancellation support when authority is withdrawn.
RESEARCH & ENGINEERING
A fabric built around
authorised, verifiable work.
Semantic networking
Information-centric networking, semantic capability discovery, and deterministic routing provide the foundation for coordinating work.
Trust & provenance
The architecture combines identity, delegated authority, policy checks, revocation, and provenance receipts. “No signature, no carriage” is its governing principle.
Efficient reuse
Inference and completed work can be candidates for reuse when their evidence, permissions, validity, and purpose remain applicable.
WHAT EXISTS TODAY
A working state-machine model; a defined programme of engineering ahead.
The current simulator executes coordination logic in Python. Providers, trust checks, identities, and receipts are modelled. Distributed enforcement, cryptographic implementation, live-system bindings, and independent security evaluation are next steps.