Meta-data associated with a dataset/file is represented using attribute/value pairs. Meta-data attributes are classified into four categories:
System-defined attributes: size (bytes), checksums (adler32, md5), creationtime, modificationtime, status, length (datasets/containers)
Physics attributes: GUID, number of events, project, datatype, run_number, stream_name, prod_step, version, campaign, lumiblocknr
Workflow management attributes: storing information like which task (task_id) or job (panda_id) produced the file
Data management attributes: necessary for the organisation of data on the grid (see Replica Management section)
For datasets, it is possible that the value of a meta-data attribute is a function of the meta-data of its constituents, e.g. the total size is the sum of the sizes of the constituents. In this case it is not possible to assign a value to it.