I created CALMS because rural schools were being excluded by “always-online” assumptions. In Zimbabwe, internet penetration is about 33%, and our IST-Africa work notes that over 70% of schools in these geolocations lack adequate digital infrastructure. CALMS therefore makes the internet optional by delivering lessons, assignments, quizzes, grading, analytics, and school operations locally over TV White Spaces, with locality-first service discovery (CNS/CIP) and deterministic routing. Design choices were validated using MyPublicWiFi simulations: bandwidth caps of 128–256 Kbps, latency of 100–150 ms and 200–300 ms, and random 5–15 minute disconnections. Under load we simulated up to 500 concurrent users; performance remained stable up to 400. Educational impact indicators improved (quiz scores 65%→78%; assignment completion +30%). Early deployments reached 500+ users across multiple schools; the current ten-school pilot pathway is structured to exceed 1,000 learners as rollout expands.
In practice, CALMS is deployed as a school-cluster system: a TV White Spaces backhaul links a local hub to participating schools, where a modest local server hosts the platform and serves phones, tablets, and laptops over LAN/Wi-Fi. Lessons and curriculum packs are pre-cached so learning continues during outages: students access content, submit assignments, take quizzes, and receive feedback; teachers upload materials, set tasks, grade, and view analytics; administrators generate reports and manage basic school operations and website/news updates. CALMS-NET (CNS + deterministic private IPv6/CIP) keeps service discovery local, reducing latency and avoiding dependence on upstream DNS/Internet. Security uses device certificates, RBAC, and end-to-end encryption; signed content packs and CNS zone bundles are pushed via scheduled multicast during off-peak hours (22:00–04:00). The model has been evaluated across ten pilot schools and stress-tested to hundreds of concurrent users, and is designed to scale to 1,000+ students by adding schools to the same TVWS-connected cluster while keeping core services local-first.
CALMS has been spreading through an iterative “pilot → refine → replicate” pathway and, increasingly, through international visibility. We started with a functional prototype focused on offline-first delivery and TV White Spaces feasibility, then piloted in a rural school with structured onboarding and feedback from 50 students and 10 teachers. Those insights drove optimisation (lighter transfers, simpler workflows, faster synchronisation, and clearer teacher/admin controls). We then replicated deployment across five additional schools, reaching 500+ users, supported by manuals/tutorials and a modular architecture that installs as a standard school-cluster kit. In parallel, our IST-Africa pathway evaluated implementation and scalability across ten pilot schools and used simulation-driven testing of multi-school clusters to confirm resilience under bandwidth limits, latency, and outages. Scaling is operationally enabled by locality-first networking (CNS/CIP), signed content packs, and scheduled off-peak multicast updates (22:00–04:00), allowing new schools to join without disrupting class time. Internationally, CALMS gained momentum through recognition as the BE OPEN “Designing Futures 2050” Public Vote winner and subsequent engagement representing BE OPEN at UNEA-7, where CALMS was presented to global stakeholders—expanding partner interest and strengthening the pathway to exceed 1,000 learners as additional schools join the cluster footprint.
We upgraded CALMS from a modular LMS prototype into a strict offline-first, school-cluster ecosystem with a clearer networking, AI, and operations stack. The IST-Africa version formalises CALMS as a four-layer architecture: (Layer 0) TV White Spaces connectivity; (Layer 1) CALMS-NET, a deterministic local routing overlay with the CALMS Name System (CNS) and a private IPv6 namespace (CIP) to keep service discovery resilient even when upstream internet and DNS are unavailable; (Layer 2) an embedded edge inference runtime (e.g., face verification and gesture recognition) so key classroom functions can run locally without cloud dependence; and (Layer 3) applications such as the e-learning hub, results management, school websites, chatroom, and the Virtual AI Classroom Board. We also added “deployment-grade” controls: device certificates, role-based access control, end-to-end encryption, signed CNS zone bundles and content packs, and scheduled multicast updates during off-peak hours (22:00–04:00) to preserve teaching-time bandwidth, plus local telemetry for basic monitoring. Finally, the evaluation approach was strengthened with more rigorous methodology and simulation testing (including fault injection and realistic link profiles), aligned to multi-school rollout.
To try CALMS, start with a small school pilot and expand to a cluster. Step 1: select 1–3 schools, appoint a focal teacher and an admin lead, and define a 4–8 week trial with clear metrics (active users, lesson completion, assessment turnaround, and uptime during outages). Step 2: prepare basic infrastructure: a modest local server/mini-PC to host CALMS, a Wi-Fi/LAN access point, and learner/teacher devices (phones, tablets, or laptops); add solar/UPS if power is unstable. Key technological components include TV White Spaces (TVWS) Base Stations and Customer Premises Equipment (CPEs) to provide reliable long-range connectivity between the hub and schools, with internet backhaul treated as optional. Step 3: install CALMS locally and preload curriculum/content packs so lessons, assignments, quizzes, grading, analytics, and school operations work offline from day one. Step 4: configure local discovery (CNS/CIP) and set roles (teacher/learner/admin); enable certificates, RBAC, and encryption for production use. Step 5: onboard teachers with short training, run weekly learning cycles, and collect feedback; use offline mode as default and sync only when backhaul is available. Step 6: scale by adding more classes and then additional schools to the same TVWS-connected footprint, using signed content packs and scheduled off-peak updates (22:00–04:00) to avoid disrupting teaching time.
