Data stored in the database mostly represent current valid state. Nowadays, complex information systems require data evolution management, state monitoring over the time. Temporal definition has been proposed, which is, however, still based on conventional paradigm. Therefore, non-current values are difficult to obtain, manage and evaluate. In this paper, several temporal evolution streams are proposed with emphasis on the historical background and technical resources and management. Own proposed temporal architecture is based on attribute; thus the granularity is shifted from object itself to column. Hybrid model provides intermediate solution between object and attribute itself and allows you to create and manage data groups. In this paper, architecture and access methods are proposed with emphasis on data flow.