DCAT distribution fields

In DCAT, a dataset record describes the intellectual object and each distribution describes one downloadable file of it. The distribution is where the operational fields live: where to fetch the file, what format it is, how big it is, under what licence, and when it last changed.

The cataloged records show both sides of the Atlantic using the same pattern. The Data.gov record description cites 'per-distribution downloadURLs and mediaType fields'; data.europa.eu is described as carrying 'a licence and byte_size recorded on every distribution' and 'per-distribution licence, byte size and modified dates'. The aluminium search record lists 'JSON-LD / DCAT-AP' among its formats, and its licence note states that most listed datasets 'display their own licence reference' while resources lacking licence information require consulting 'the licence conditions in the original portal'. That last clause is the operational catch: the field can be empty, and empty means a trip to the source.

Distribution fields are the difference between a harvest script that survives and one that downloads 4 GB of PDFs at 3 a.m.

• mediaType prevents format surprises. A dataset advertising CSV may distribute XLSX; checking mediaType per distribution catches that before your parser breaks.

• byte_size is your bandwidth budget. data.europa.eu records byte size on every distribution, so a harvest can skip or schedule the multi-gigabyte files instead of discovering them mid-download.

• Licence per distribution is a legal boundary. Because the licence URI attaches to each file, one dataset can mix commercial delivery terms 4.0 and unlicensed resources - redistributing the bundle without reading each licence is the risk.

• Modified dates drive incremental pulls. Per-distribution modified dates let you re-fetch only changed files; dataset-level dates hide which of twenty distributions actually moved.

1. Confirm downloadURL and mediaType exist per file. The cataloged Data.gov description cites 'per-distribution downloadURLs and mediaType fields' - a catalog that only links landing pages cannot be harvested automatically.
2. Check byte_size before large pulls. data.europa.eu records 'a licence and byte_size recorded on every distribution'; use it to schedule or skip heavy files.
3. Read the licence URI on every distribution. The licence is 'declared per distribution via a URI (e.g. creativecommons.org/licenses/by/4.0)'; treat a missing licence as unresolved and consult the original portal, as the cataloged terms direct.

Terms that sit next to distribution fields in a DCAT harvest:

• dcat-catalog-search-api — The unauthenticated search endpoint that returns records with these distribution fields.
• per-distribution-license — The licence declaration attached to each downloadable file rather than the dataset.