File-Based Integration
Exchanging data as files on a schedule via SFTP or shared storage — dated, still ubiquitous, and workable when its failure modes are handled explicitly.
Still the dominant integration mechanism in banking, insurance, healthcare and government, because it requires nothing of the legacy system beyond producing a file.
Its failure modes are well known and each has a standard remedy. Partial reads — a consumer reading a file while it is still being written — are solved by writing to a temporary name and atomically renaming, or by a marker file written after the data. Duplicate processing needs an idempotency record of processed filenames and checksums. Missing files need an expected-arrival alert, because the absence of a file is silent. Out-of-order arrival needs sequence numbers rather than timestamps.
The properties that keep it viable: files are auditable, replayable, and inspectable by non-engineers, which is why finance teams often prefer them.
Modernise it by wrapping rather than replacing — land the file in object storage, emit an event on arrival, and process it through the normal pipeline.