Something I noticed whilst comparing the postgres and mysql rlm_sql configs is that when postgres sets an AcctStartTime or AcctStopTime, it figures in the delay time, whereas the mysql queries don’t.

For example, in accounting_stop_query, postgres has:

AcctStopTime = (now() - '%{Acct-Delay-Time:-0}'::interval)

… whereas mysql just has:

AcctStopTime       = '%S'

Where %S is “request timestamp in SQL format”.  Which should be pretty darn close to now().

I presume this is just down to the personal preferences of whoever wrote the two sets of queries?  Personally, I wouldn’t expect the queries to apply the delay time to the start/stop time, so the MySQL version is “correct”.  IMHO interpretation of the data should be left to the backend provisioning.

No biggie, just an observation.

   -- hugh