Thank you all for the information and ideas, however something doesn't fit here ... or it could just be my math :) ======================================================================= This is from my host ... -> uname -a Linux myhost 2.6.18-238.9.1.0.1.el5 #1 SMP Tue Apr 12 19:58:35 EDT 2011 x86_64 x86_64 x86_64 GNU/Linux ->last reboot reboot system boot 2.6.18-238.9.1.0 Wed Jun 26 17:12 (57+22:15) reboot system boot 2.6.18-238.9.1.0 Wed Jun 26 16:49 (00:22) wtmp begins Wed Jun 26 16:49:17 2013 ======================================================================= These are beginning lines from a simple 10046 trace file *** 2013-08-23 15:24:25.823 *** SESSION ID:(153.12935) 2013-08-23 15:24:25.823 *** CLIENT ID:() 2013-08-23 15:24:25.823 *** SERVICE NAME:(SYS$USERS) 2013-08-23 15:24:25.823 *** MODULE NAME:(sqlplus@somehost.somewhere (TNS V1-V3)) 2013-08-23 15:24:25.823 *** ACTION NAME:() 2013-08-23 15:24:25.823 WAIT #47561771746056: nam='SQL*Net message to client' ela= 2 driver id=1650815232 #bytes=1 p3=0 obj#=-1 tim=1377285865823053 << lets use this *** 2013-08-23 15:24:31.018 WAIT #47561771746056: nam='SQL*Net message from client' ela= 5195574 driver id=1650815232 #bytes=1 p3=0 obj#=-1 tim=1377285871018953 CLOSE #47561771746056:c=0,e=13,dep=0,type=1,tim=1377285871019108 ======================================================================= Now I'll take a tim value and try to convert it alter session set nls_date_format='mm/dd/yyyy hh24:mi:ss'; set numwidth 20 line 120 with x as (select 1377285865823053 as t, (1377285865823053/1e6) as t1 from dual) select t tim_micro, t1 tim_sec, sysdate curr_sysdate, sysdate-(t1/86400) derived_sysdate from x; TIM_MICRO TIM_SEC CURR_SYSDATE DERIVED_SYSDATE -------------------- -------------------- --------------------- --------------------- 1377285865823053 1377285865.823053 08/23/2013 15:29:42 12/31/1969 20:05:16 something doesn't match up, tim is neither time from epoch, not it is from last system reboot on my system or why bother ? even if i factor GMT it is still off by over 5 minutes. Raj On Fri, Aug 23, 2013 at 12:24 PM, Niall Litchfield < niall.litchfield@xxxxxxxxx> wrote: > Raj > > At least on Linux (but probably everywhere *nix) the times come from the > gettimeofday system call. Last time I looked on Windows it was time since > last boot (but that was no later than 9i and might have been 8i). > gettimeofday should be documented but from memory is Jan 1 1970 > -- //www.freelists.org/webpage/oracle-l