Offline Stack
Systems That Keep Working When the Network Doesn't.
In progress
When the uplink dies, most modern infrastructure dies with it. This book treats offline operation as a property of an entire dependency graph: a capability like navigation, reference, communication, or recovery survives only if every dependency below it does. The chapters cover knowledge libraries built on Kiwix, ZIM, and offline maps; ISO, git, OCI, and package mirrors that can rebuild systems from scratch; local services from DNS and NTP to LLMs; links from LoRa mesh to packet radio and sneakernet; and the power budgets and drills that prove the stack works with the WAN unplugged.
Status: the manuscript baseline is complete. All 30 chapters and 6 appendices are drafted, and fact-checking is underway.
The first edition will ship DRM-free as PDF and EPUB, with free updates.
Capability by capability, plane by plane.
30 chapters and 6 appendices across seven parts, following one reference design from dependency closure to a 72-hour disconnection exercise. Chapter titles may still shift during fact-checking and revision.
Part 1 Failure is a design input 4 chapters
- 1 Offline is a system property
- 2 Missions, modes, and service levels
- 3 Reference architecture
- 4 Inventory and dependency closure
Part 2 Knowledge that survives 5 chapters
- 5 Library architecture
- 6 Curating an operational library
- 7 Kiwix, ZIM, and web archives
- 8 Offline maps and navigation
- 9 Private records and field copies
Part 3 Software that can rebuild itself 5 chapters
- 10 The artifact depot
- 11 The git source ark
- 12 OCI and package mirrors
- 13 Disconnected development
- 14 Integrity, backup, and restore
Part 4 Services close to home 4 chapters
- 15 The local network control plane
- 16 Media, learning, and collaboration
- 17 Local LLMs and retrieval
- 18 Field nodes and the portable stack
Part 5 Beyond conventional IP 4 chapters
- 19 Constrained-link engineering
- 20 LoRa, Meshtastic, and MeshCore
- 21 Packet radio and amateur IP
- 22 Delay-tolerant networking and sneakernet
Part 6 Power, recovery, and routine 4 chapters
- 23 Power budgets and UPS design
- 24 Solar and DC power
- 25 Contingency, security, and drills
- 26 Spares, repair, and maintenance
Part 7 Infrastructure people can share 4 chapters
- 27 Serving neighbors and local groups
- 28 Governance, safety, and law
- 29 Tanaka extended reference design
- 30 The 72-hour disconnection exercise
Appendices Appendices 6 appendices
- A Capability-based bills of material
- B Content manifest and refresh matrix
- C Service cards, runbooks, and drill cards
- D Link planning and radio-law checklist
- E Sources, reproducibility, and verification
- F Glossary and subject index
For engineers who want their systems to outlast their uplink.
This book is for sysadmins, SREs, developers, digital preppers, and radio experimenters who want a working local stack rather than a shelf of unopened gear. It assumes comfort with Linux and a command line, and it evaluates every design in connected, WAN-isolated, power-constrained, and field modes.
- Sysadmins and SREs who run home labs or community infrastructure and want services that keep answering while the WAN is down.
- Developers and data hoarders who mirror ISO images, git repositories, OCI images, packages, and models, and want those mirrors verifiable and restorable.
- Digital preppers and radio experimenters who build LoRa mesh, packet radio, and sneakernet links and need them to carry real services within legal limits.
In the works.
The manuscript is drafted and moving through fact-checking. Join the shared Sysinit Press book list and you'll get one message when it ships, with a sample chapter as soon as one is public. The first edition will include DRM-free PDF and EPUB files with free updates.
The signup uses double opt-in. Every message includes an unsubscribe link.