From our specification:

"The JSON formatted incremental data dumps which only contain those documents which have changed since the last incremental data dump, are located on the HTTPS secured domain metabrainz.org . To access these files, an access token and a replication sequence number are required. The replication sequence number is a monotonically increasing number that will be incremented for each partial dump. Licensee can generate the access token after this contract is executed; this access token must be supplied when fetching the incremental data dumps and must follow the the URL structure shown below. The access token can remain constant over time and does not need to be renewed periodically. For each of the Entities the name of the entity and the access token can be substituted into the URL below to fetch that data dump:

https://metabrainz.org/api/musicbrainz/json-dumps/json-dump-<replication seq>/<entity>.tar.xz?token=<access_token>"

The partial data dumps will follow the same file naming structure as the full json dumps:

  ftp://serge.metabrainz.org/pub/musicbrainz/data/json-dumps/20170523-203331/

There will be a partial dump file for each MB entity, except URL and recording.