[ https://issues.jboss.org/browse/TEIID-5290?page=com.atlassian.jira.plugin... ] Steven Hawkins commented on TEIID-5290: --------------------------------------- This is mostly about the function definition for now/current_timestamp - it is defined as command deterministic, which as per TEIID-5204 means that it should have a single value for the scope of the user command. There is no difference in the behavior in this case between using "declare timestamp ts2 = now()" and "declare timestamp ts2 = (select now())" as we are still only looking for the command level value. getdate and clock_timestamp are the fully non-deterministic timestamp functions for sql server and postgresql respectively. Postgresql actually offers: transaction_timestamp() - same as pg now() statement_timestamp() clock_timestamp() Our current behavior for now / current_timestamp matches pg statement_timestamp. If you use now() in the postgresql example above, it will produce the same value for each timestamp. It appears that SQL Server and postgresql disagree on the meaning of current_timestamp - it's fully non-deterministc for SQL Server and it matches transaction_timetstamp for postgresql. The spec does not seem to fully specify what current means. Our choices here are: 1. leave the Teiid current_timestamp / now functions as command deterministic as per TEIID-5204 and introduce a new fully non-deterministic timestamp function. 2. change the Teiid current_timestamp / now functions back to the non-determinisic implementation and update the metadata to indicate it is fully non-deterministic. Do you have a preference as to the resolution? -- This message was sent by Atlassian JIRA (v7.5.0#75005)