The general consistency requirement for a high performance database management system (DBMS) are stated. The CODASYL Database Task Group (DBTG) proposals are examined in the light of these requirements and are shown to be inadequate in terms of integrity safeguards, where too great a burden is placed upon the applications programmer, and in terms of support for concurrent run units, where the inherent performance constraints are likely to embarrass rapid response systems. Some revisions to the CODASYL DBTG proposals are put forward which would remove these inadequacies.