Université libre de Bruxelles API Rules

name: DI-fusion Consumption Rules
aid: ulb:difusion-export
generated: '2026-08-30'
method: derived
source: >-
  Derived from openapi/ulb-difusion-export-openapi.yml, openapi/ulb-difusion-oai-pmh-openapi.yml,
  errors/ulb-difusion-errors.yml and the DI-fusion terms of use at
  https://bib.ulb.be/en/find-documents/di-fusion/terms-of-use (HTTP 200, 2026-08-30). These are
  rules for CONSUMING ULB's surfaces safely, not a lint ruleset for the specification.
x-operator: institution
rules:
  - id: difusion-no-4xx
    severity: error
    statement: >-
      Never treat a 5xx from difusion-svc.ulb.ac.be as a transient outage. The service returns
      HTTP 500 for missing mandatory parameters and for unknown scholar identifiers, so a retry
      loop on 500 will hammer the host and never succeed. Validate scholarID and scholarInst
      client-side before the call.
    evidence: errors/ulb-difusion-errors.yml
  - id: difusion-reftype-allowlist
    severity: error
    statement: >-
      Request only reftype values verified reachable — xml-brief, apa, bibtex, ris, csv. xml-full
      returns 403 and xml-brief-ext returns 500, despite both being documented.
    evidence: errors/ulb-difusion-errors.yml
  - id: difusion-reftype-filetype-pairing
    severity: error
    statement: >-
      Send filetype only in a pair the service permits; the allowed combinations are enumerated in
      vocabulary/ulb-difusion-vocabulary.yml under reftype-filetype. csv accepts only csv, and the
      xml reftypes accept only xml.
    evidence: vocabulary/ulb-difusion-vocabulary.yml
  - id: difusion-markup-scope
    severity: warn
    statement: >-
      The markup letters b and o exist on the scholar operation only, and only take effect for
      certain reftype/filetype pairs. Sending an ineffective markup letter is silently ignored, so
      do not infer support from a 200.
    evidence: vocabulary/ulb-difusion-vocabulary.yml
  - id: difusion-solr-encoding
    severity: error
    statement: >-
      The query parameter is passed through to Apache Solr. A literal colon must be escaped as
      `\:` and the whole value URL-encoded, or the filter silently changes meaning.
    evidence: openapi/ulb-difusion-export-openapi.yml
  - id: difusion-xml-not-json
    severity: warn
    statement: >-
      Every non-binary response is XML or plain text with charset=UTF-8. There is no JSON
      representation and no content negotiation; Accept headers are ignored. Parse the element
      tree described in json-schema/ulb-difusion-publist-schema.json.
    evidence: json-schema/ulb-difusion-publist-schema.json
  - id: difusion-arrays-always
    severity: warn
    statement: >-
      Every record field arrives as an `<arr name="...">` list even when single-valued. Treat all
      of them as arrays; a consumer that assumes a scalar will break on the first multi-author
      record.
    evidence: examples/ulb-difusion-scholar-xml-brief.xml
  - id: difusion-conditional-fields
    severity: warn
    statement: >-
      abstract appears only under markup=a and the three ftlink_* fields only under markup=l. Their
      absence is a request-shape consequence, never a statement that the repository lacks the data.
    evidence: json-schema/ulb-difusion-publist-schema.json
  - id: difusion-fulltext-host-unstable
    severity: warn
    statement: >-
      ftlink_* values point at https://dipot.ulb.ac.be/dspace/bitstream/..., and dipot.ulb.ac.be
      was returning "DSpace at ULB: Internal system error" under HTTP 500 on 2026-08-30. Do not
      assume a returned full-text link resolves; check it.
    evidence: errors/ulb-difusion-errors.yml
  - id: oai-do-not-trust-registration
    severity: error
    statement: >-
      DI-fusion is registered as OAI-PMH compliant in OpenDOAR, ROAR and Sherpa, and its endpoint
      answers Identify and ListMetadataFormats. It still harvests nothing — every record-bearing
      verb rejects the metadataPrefix it advertises. Verify with an actual ListRecords before
      scheduling a harvest.
    evidence: errors/ulb-difusion-errors.yml
  - id: oai-parse-before-status
    severity: warn
    statement: >-
      The OAI-PMH endpoint returns HTTP 500 alongside a valid protocol document, with HTML appended
      after </OAI-PMH>. A harvester must truncate at the closing tag and must not gate on the status
      code, or it will discard a correct Identify response.
    evidence: examples/ulb-difusion-oai-identify.xml
  - id: oai-ignore-echoed-baseurl
    severity: warn
    statement: >-
      Do not follow the `<request>` value in any response from this endpoint. It echoes the
      unmodified VuFind demo default http://digital.library.villanova.edu/OAIServer.php, which is
      Villanova's host, not ULB's.
    evidence: examples/ulb-difusion-oai-identify.xml
  - id: difusion-attribution
    severity: error
    statement: >-
      The DI-fusion terms of use permit harvesting for non-profit purposes provided the OAI
      identifier or the source link is indicated. Carry startpage or the OAI identifier through to
      any derived product.
    evidence: https://bib.ulb.be/en/find-documents/di-fusion/terms-of-use
  - id: difusion-personal-data
    severity: error
    statement: >-
      Publication lists are keyed on an ULB matricule — an internal person identifier — and resolve
      to a named individual. ULB is a Belgian public institution under GDPR. Do not build or
      republish a matricule-to-person index from bulk enumeration of scholarID.
    evidence: https://www.ulb.be/fr/mentions-legales/politique-de-protection-des-donnees-a-lulb
  - id: difusion-no-rate-limit-published
    severity: warn
    statement: >-
      No rate-limit, quota or Retry-After header is present on any response, and ULB publishes no
      usage policy for the export service. Absence of a published limit is not permission for
      volume; pace requests and identify your client.
    evidence: authentication/ulb-authentication.yml