Key takeaway
Identify the representation and access route behind each catalog entry. Two formats can describe one dataset, and an access URL can lead to a page or service.
Identify what the catalog node represents
A buyer receives three catalog links: a CSV, a JSON representation and an API page. Treating them as three independent datasets can inflate both a coverage claim and a delivery plan. Start by asking which underlying resource each link describes and how the intended representation can actually be obtained.
W3C DCAT 3 distinguishes a Dataset, a Distribution and a Data Service. Its accessURL can identify a location or service giving access, while downloadURL identifies a downloadable file. A valid catalog entry therefore need not expose a direct file. These are metadata meanings, not proof of availability, quality or permission for the buyer’s purpose.
Resolve one fictional archive into its access objects
Hypothetical example: a catalog describes maintenance dataset D7, with CSV distribution C7 and JSON distribution J7 containing the same observations. A service S7 provides an approved query route over D7. The names are invented labels, not real catalog resources or stocked VOID inventory.
The completed map keeps one underlying dataset distinct from its representations and service. It also keeps descriptive metadata separate from a proposed recipient’s authorization. A public description can be read while any actual data access remains withheld.
| Invented node | Meaning in this example | Buyer’s next check |
|---|---|---|
| D7 | Underlying maintenance dataset | Confirm covered observations, period and source |
| C7 | CSV representation of D7 | Locate direct file if offered; verify format and version |
| J7 | JSON representation of the same D7 observations | Do not count it as additional independent coverage |
| S7 | Service providing access to D7 | Inspect endpoint description, permitted query and access conditions |
| Catalog access page | Route explaining how access works | Do not treat its HTML as the data file |
Keep the file claim on the file representation
DCAT places format, byte size and download information on the distribution. The European Commission’s DCAT-AP 3.0.1 profile similarly distinguishes an access route from a direct download and relates a checksum to the downloadURL. Its profile-specific requirements should not be relabeled as universal requirements for every DCAT implementation.
For C7, the evaluator asks which exact CSV representation is proposed and where its checksum and schema apply. For J7, a different byte count or checksum would be unsurprising even if the observations were equivalent. Byte equality and informational coverage are different comparisons.
If the API returns a filtered period, record the actual request, scope and resulting artifact. The existence of S7 does not establish that the full D7 archive was retrieved. Pagination, query limits and later source updates require their own completeness and identity evidence.
Read the rights statement at the relevant level
DCAT 3 recommends license and rights information at distribution level and warns against conflicting dataset/distribution statements. That helps an evaluator find the relevant information. The metadata does not itself decide whether the publisher has the rights it describes or whether the proposed use satisfies applicable conditions.
For the fictional record, a dataset-level label says evaluation access, while an old C7 description says unrestricted reuse. The buyer records that inconsistency and holds the proposed use pending clarification of the actual terms. Choosing the more permissive text would not resolve the conflict.
Keep authentication instructions, public metadata and permission evidence separate. An endpoint responding successfully does not grant a new training purpose. Conversely, an approved purpose does not prove that a documented download route works in the recipient’s authorized environment.
Accept a defined access artifact
The completed buyer request names D7, the intended C7 representation, its covered period, format/schema reference, access method and unresolved rights issue. The three original links become one scoped delivery inquiry. No source count or usable-coverage claim is inferred merely from the number of links.
Use due diligence to request those distinctions and inventory to retain the source description. VOID helps explore a potential fit and a permissioned named introduction. It does not operate the fictional catalog, supply an API credential or offer the described archive for download.