This metadata can be extracted programmatically and written into database columns as the data is inserted. – and some metadata that can be added by the author. Text formats (PDF, Microsoft Word) generally contain some automatically generated metadata – the author's name, document title, last modified date, etc. Some formats have metadata embedded in them. There are several ways to create the metadata: Then we can query the metadata to find a particular instance of a LOB or to find out information about an instance of a LOB. Once we have the data in a LOB, we can store it in a database table and add metadata in other columns in the table. But the data is opaque i.e., nothing is known about a LOB instance except that it may be large.

A LOB is handled in a special way by the database to cope with its potentially large size – typically two or four gigabytes. A binary LOB is a LOB that contains data that is not character-based, as opposed to a character LOB, or CLOB, which contains data that is in the character set of the database.

LOB storage just means that the data item is stored as a single item in one place. Despite its name, a LOB has none of the useful attributes of an object. In a database this opaque chunk is often called a BLOB, or binary large object. One approach is to store the nontraditional data as an opaque chunk of data and add metadata.

Jim Melton, Stephen Buxton, in Querying XML, 2006 2.3.1 Metadata

Metaclean full