Which two statements are true about Oracle Data Pump export and import operations?

Data Pump export runs as a persistent master control process that continues independently of the initiating client; the client can exit and later reconnect using the EXPDP ATTACH=job_name parameter, allowing long exports to run unattended without losing job state.

When the PARALLEL parameter is specified for an import, Data Pump spawns Oracle parallel execution (PX) server processes that simultaneously load data into separate table segments or partitions, directly reducing total import elapsed time for large data sets.

Data Pump file access is governed entirely by Oracle DIRECTORY objects pointing to OS paths; the underlying directory only needs to be accessible to the Oracle process - there is no requirement that it be owned by the oracle OS account.

Data Pump selects the load or unload method per object type; referential integrity constraints, bitmap indexes, and certain data types force use of the conventional path method, so direct path is not applied universally to all object types.

Unlike export jobs, Data Pump import jobs do not support the ATTACH parameter for reconnecting a detached client session; once the initiating session ends, the client cannot reattach to regain interactive control of the import job.