According to the protocol, each record contains a header with a datestamp with "the date of creation, modification or deletion of the record for the purpose of selective harvesting."
The protocol also explains the selective harvesting as follows:
- "..modification - the response must include records, corresponding to the metadataPrefix argument, which have changed within the bounds of the from and until arguments
- creation - the response must include records, corresponding to themetadataPrefix argument, that have become available from the repository within the bounds of the from and until arguments
- deletion - depending on the level at which a repository keeps track of deleted records, the response may include headers of records, corresponding to the metadataPrefix argument, which have been withdrawn from the repository within the bounds of the from and until arguments. Deleted status is indicated via the status attribute of the header element and no metadata is included..."?
It is very, very important to take great care in implementing the datestamp according to the protocol specifications as quoted above. Experience has taught that many harvesting errors that occur with incremental harvesting have their origin in misinterpretation of the datestamp.