|
|
|
Consider the following when implementing a search strategy:
Where possible, connectors should use filtering restrictions at the lowest possible level in the implementation. Filtering restrictions prevent extra search results from being returned which you then filter out at higher levels at greater performance cost. Connectors must at minimum support filtering based on object class, and against naming attribute values (both exact and wildcard matching). The Java CS can, optionally, as configured through metadata, deal with performing post-filtering on search results. This accounts for cases where the connector cannot natively respect all the clauses in a search filter.
The ability to stream large sets of search results is an important behavioral characteristic for a connector. While it is possible to achieve streaming with most APIs, a performance cost can be incurred. Where streaming cannot be achieved (for example, with DYN or JDBC), do the following: