Introducing Favonius

Most organisations do not have a shortage of places to keep data. They have a shortage of dependable ways to move it.

A production build needs to reach a remote site. A media team needs to deliver source material to a partner. An analytics group needs a current data set in a different cloud region. A recovery workflow needs to copy a large collection of files before its operating window closes. In each case, the transfer is part of a business process, not an incidental command-line task.

Today we are releasing Favonius 0.1, an Apache-2.0 open-source file-transfer and synchronisation engine for moving large files and directory trees across long, variable, or lossy network paths.

It uses a purpose-built protocol over UDP, with adaptive congestion control, so that a transfer can make better use of the capacity available on paths where conventional transfers often become unpredictable. Explore Favonius.

Resume rather than restart. Interrupted transfers compare the source and destination with BLAKE3 Merkle trees and send only missing or changed chunks. A stalled multi-gigabyte transfer should not automatically become a full retransmission.

Move trees, then keep them current. Favonius transfers directory trees with include and exclude rules, and runs stateless one-way, mirror, or append-only synchronisation. This supports repeatable hand-offs, replication jobs, and data staging.

Protect data in transit. Optional end-to-end encryption uses modern cryptography, with key rotation and optional pinned server authentication. The receiving daemon can be confined to a specified destination root, preventing a sender from choosing arbitrary paths on the host.

Work with the infrastructure already in place. Favonius supports IPv4 and IPv6, can compress suitable content per chunk, and exposes Prometheus metrics for transfer activity, bytes moved, packet counts, and latency.

Conventional file-transfer tools remain excellent choices for short, clean networks. But latency and intermittent loss change the economics of a transfer: available bandwidth may be ample while the application still takes far longer than its operating window allows.

Favonius continuously paces traffic and offers congestion-control profiles for different path conditions. The default makes a conservative choice for wide-area paths, while operators can select a profile for a known environment. The aim is straightforward: make delivery more predictable without asking a team to tune every transfer by hand.

A shorter, more reliable transfer window can mean fresher data for a regional operation, less overnight capacity reserved for replication, and fewer manual interventions after a connection drops. It also lets teams treat data movement as an observable service with explicit security and destination boundaries, rather than as an opaque background job.

Favonius is a transfer engine, not a complete file-sharing or data-governance platform. It is a good fit when an engineering or operations team needs a self-hosted, scriptable component for bulk delivery and synchronisation. It can sit behind an existing identity, scheduling, storage, or workflow layer.

Favonius is written in Rust and ships as a command-line sender with a receiving daemon. The repository includes a local quick start, deployment guidance, security notes, protocol documentation, and benchmark harnesses.

Favonius 0.1 is an early release. Organisations should evaluate it in a controlled environment, review the security guidance, and plan for protocol changes during the 0.x series. If your work depends on moving large or frequently changing data across real networks, we would welcome your evaluation and feedback at vtn.li/favonius.