OAI-PMH for scholarly journals: an implementation-minded guide

What OAI-PMH does for a journal

The Open Archives Initiative Protocol for Metadata Harvesting provides a standard way for a data provider to expose metadata and for service providers to harvest it. It does not publish the article, assign a DOI or guarantee inclusion in an index. Its value is interoperability: a journal can make consistent records available without building a separate custom feed for every harvesting service. The public article page remains the human-facing record while OAI-PMH supplies a machine-readable route to its metadata.

Understand the six protocol requests

A conforming repository responds to Identify, ListMetadataFormats, ListSets, ListIdentifiers, ListRecords and GetRecord. Identify describes the repository; ListMetadataFormats declares available schemas; ListIdentifiers and ListRecords support harvesting; GetRecord returns one record; and ListSets enables optional grouping. Requests use one base URL and responses are well-formed UTF-8 XML. Supporting the protocol means handling valid requests, required arguments and defined error responses, not merely returning an XML page.

Identifiers and datestamps control reliable harvesting

Each item needs a persistent OAI identifier that does not change when descriptive metadata is corrected. The record datestamp represents creation, modification or deletion for harvesting purposes and allows incremental collection. Confusing it with the article’s original publication date can cause updates to be missed. When a record is withdrawn, the repository’s deletion policy and response behavior should be deliberate so harvesters can reconcile their copies.

Pagination, sets and metadata formats need testing

Large result lists use resumption tokens. Tokens must preserve the original query context, expire predictably if expiry is used and avoid returning duplicates or gaps as the archive changes. Sets can expose journals, collections or publication groupings, but unnecessary complexity makes harvesting harder. At minimum, OAI-PMH requires Dublin Core support; additional formats should only be advertised when every response validates against the declared schema.

Operate the endpoint as a production service

Validation should cover XML schema conformance, character encoding, date ranges, identifier stability, deleted records, token flow and representative metadata. Monitor errors and response time, use compression where appropriate and protect the application from abusive request rates without blocking legitimate harvesters. Dijider can generate OAI-PMH output from the same controlled article record used for public pages and other exports. That shared source reduces contradictions, but ongoing metadata correction and endpoint monitoring remain essential.