Research record from 6 November 2026. This is an evidence-led account of the research and design, not an announcement of independently audited software.
Commonveil's offline research separates implemented encrypted queues and local sites from a future native Bluetooth and Wi-Fi mesh design.
Why unreliable connectivity is a design requirement
A private communications network should not become unusable whenever a mobile connection disappears. The Commonveil investigation looked at three distinct problems: writing a message while the browser is offline, transferring it between sites after a temporary federation outage, and physically carrying sealed messages between nearby devices without the internet.
The first two have documented implementations. The third remains a design for a future native application. Treating all three as a finished peer-to-peer mesh would erase a major technical distinction.
What the browser can do today
The offline outbox seals and signs messages when the user writes them, before they are stored locally. It uses IndexedDB to retain encrypted requests rather than saving plaintext messages. When connectivity returns, queued sends are retried in order with stable message identifiers so a repeated delivery does not create duplicate messages.
The design also allows messages written within a defined time window to retain their original timestamps. If group membership or keys changed while the sender was away, the old sealed request can fail safely. The interface then needs to offer a clear option to reseal and resend or delete, rather than silently sending with outdated security assumptions.
Why a local site is different from a mesh
A Commonveil site on a Raspberry Pi or other local server can serve members on the same network without a route to the wider internet. Users still communicate with a server, but it is their chosen server. Sites can pass queued encrypted requests once another site becomes reachable.
An actual phone-to-phone mesh would work differently. Devices would discover neighbours and carry sealed requests across short-range links. Modern web browsers cannot reliably perform the necessary background Bluetooth and Wi-Fi Direct operations on ordinary phones. That constraint is why a native app is part of the documented future design.
The transport research and trade-offs
The mesh design proposes store, carry and forward rather than building a blockchain. A relay device should not need to inspect message content or know the sender's account. Rotating identifiers, hop limits, expiry rules and per-sender budgets can narrow tracking and abuse risks.
Store-and-forward also has unavoidable limits. A message may not find a route before expiry. A carrier can drop traffic. The transport can expose destination-site information and message sizes even while the content remains encrypted. A UI must report pending delivery honestly.
What has been tested and what has not
The repository documents tests for offline delivery, deduplication and temporary cross-site holding. The native mesh itself is not implemented and should not be advertised as available. The next engineering work would require a native client, nearby-device exchange, storage budgets and an endpoint able to accept carried batches.
The useful conclusion is that encryption should be independent of transport. That allows one correctly sealed message format to travel online, across local networks or eventually by an opportunistic mesh without reinventing identity and message protection.
Primary sources and further reading
Explore other Comutability research ยท Read about the organisation
