SRU enables you to conduct purposeful research without having your own database.
OAI facilitates the continual synchronisation of large amounts of data. This requires the import of current basic stock into a separate database.

An OAI request returns all data records that were newly created or modified during the period defined. Retrospective harvesting over several months is not recommended, as certain automated processes have been put in place to ensure that the data records are continually updated. This means that any data harvested retrospectively would be incomplete. We provide complete sets of any retrospective data required.

OAI cannot be used to request data records using criteria other than the date on which they were changed. Please use the SRU interface for such requests.
The OAI interface cannot be used to request all new data records. Why not?
When a new data record is created, the change date is the same as the creation date. If this record is changed manually or automatically at a later date, the change date is adjusted and a request sent to the OAI interface specifying the period in which the record was created will return no hits even though the record was actually created at this time. The record will only be delivered if the time of the last change is specified, as this is the only criterion recognised by OAI.

The German National Library advises users not to perform searches of the repository at intervals of less than one minute, since data may otherwise be delivered in duplicate (does not apply to OAI requests using “resumptionToken”). The period between searches should not be shorter than the defined period.

Hit lists are limited to 100,000 data records each. An error message is returned if the number of hits is greater. If this occurs, the search period and frequency must be reduced accordingly.

A limited search period should be defined in order to ensure that not more than 100,000 hits are returned. Recommended search period/frequency for searches that are not time-critical: 30 minutes. In the case of smaller sets (e.g. online dissertations), harvesting can be restricted to once a day or once a week, since data records changed several times during this period will only be harvested once to prevent too many hits from being returned.
We also recommend using the time setting in the “responseDate” element, e.g. <responseDate>2017-08-30T08:12:54Z</responseDate> as the retrieval time (“from”), as this time most accurately reflects the current availability of data in our repository. We also recommend harvesting with a small time overlap (“responseDate” minus one minute = “from”).

In the case of special operations requiring changes to more than 50,000 data records in one day, we proceed as follows:

• Extensive special operations are carried out at weekends/on public holidays.
• The changes to the data records are not logged, which means that this data is not visible via the OAI index.
  The changes to the data records are visible only in the OAI index at a later date when the OAI index has been completely newly created.
  
• OAI deletions are not delivered.

A system that harvests data promptly after a change will therefore miss the changes, even if it running continually. Only if the changes are relevant to the harvesting system can the harvesting interval be reset as necessary.

OAI deletions are not delivered.
For ZDB bibliographic and local data and the GND, deleted and redirected records are delivered as truncated data records and labelled as such; they are furnished with ID numbers, standard and date fields.
Local data marked as deleted is available via the OAI interface for no more than one week (further information in FAQ on ILTIS).
The process for title changes has not yet been finalised.

Only the data records for each part (with dependent or independent title) are delivered via the OAI interface. In contrast, the data service delivers both the data records for each part and the superordinate data records via SFTP/WWW servers. In general, OAI-PMH can be used to retrieve the headers by entering the corresponding identifiers. Depending on the type of data record (with dependent or independent title) and format, this identifier is stored in various fields or elements of the data records for each part, cf. overview (available only in German).

No, because data records are constantly being changed and can only be searched via OAI using the changed date, which means that it is impossible to predict whether a particular record will be contained in an OAI response.

A “resumptionToken” facilitates the delivery of partial responses when a search returns 51 or more data records. The OAI harvester receives a token which it has to use for an additional request from the OAI repository in order to get the next partial response. If not all records have yet been delivered, a new “resumptionToken” is generated with each partial response and must be used to obtain the next. The element “resumptionToken” contains the attribute values of the current list position and the total number of data records (cursor="50" completeListSize="xxxxxx"). Each OAI response delivers a maximum of 50 data records. A “resumptionToken” is valid for no more than 30 minutes.

Example using a “resumptionToken”:

https://services.dnb.de/oai/repository?verb=ListRecords&resumptionToken=xxxxxxxxxx_hier_resumptionToken_einfügen_xxxxxxxx

UTC (Universal Time Coordinated) is used as the basis for calculating the local time.
Example for Germany:
Local time for Berlin is UTC+1 (during summer time it is UTC+2). The OAI server gives the time in UTC.
If the local time is 10:30, it therefore only makes sense to harvest until UTC 09:29, or until UTC 08:29 in summer time.

Mark-up characters (<, > and &) in an OAI response are transported in a CDATA section.

Please send any specific queries or error messages to the OAI interface without delay. This is the only way to ensure transparency and help resolve the problem.
Please send queries or error messages to schnittstellen-service@dnb.de together with the following information:

• Syntax of OAI-PMH request (set, format, time period)
• Error message/description
• Context (repeated, sporadic, client used etc.)

We recommend setting the time-out to at least 2 minutes so that the harvesting process is not interrupted. If extensive changes are made, the response times for certain sets (e.g. zdb:holdings) may last more than 90 seconds even if short query times are entered.

If you are just getting started with processing metadata, useful programmes include the software suite Catmandu, OpenRefine or Metafacture, while data can be analysed with "Konstanz Information Miner" (KNIME) or the Metadata Quality Assurance Framework. A more detailed overview is provided in the presentation slides "Open Source Software zur Verarbeitung und Analyse von Metadaten" (available only in German) and the article "Survey of Tools for Linked Data Consumption".

We suggest you sign up to the OAI interface mailing list to make sure you receive timely information about changes, new developments, disruptions and maintenance work.