Re: MIDRANGE-L Digest, Vol 3, Issue 1010

[Tom Liotta <qsrvbas@xxxxxxxxxxxx>]

midrange-l-request@xxxxxxxxxxxx wrote:

>   6. RE: Fastest way to get a unique identifier/tracking column
>      changes (Reeve)
>
> I don't believe the results will be different depending upon which approach,
> trigger or journal receiver, you use.  It's about implementation, stability,
> and performance.

I'd suspect this to be more or less true depending on the purpose. If 
it's simply a form of historical tracking that perhaps supports 
drill-down or other application functions, then a trigger writing to a 
file is perfectly reasonable. But if this is an actual audit function 
mandated by auditing requirements or regulations, then journaling is the 
only reasonable choice.

Journal entries are system-enforced and reasonably secure from 
tampering. Logging to a physical file provides no significantly secure 
audit capability at all. Anybody with authority can generate rows, 
update or delete rows, and put any desired info into or take undesired 
info out of a physical file. But manipulating journal entries is 
significantly more difficult.

There's no way to demonstrate that the rows in a log file were the 
result of actual operations -- unless, of course, you also journal 
activity against the log file which brings you right back to journaling 
and only demonstrates activity against the log file not the original 
tables. IMO, as audit evidence, a log file is effectively worthless. 
Opinion? Yes, certainly. But I haven't seen a good way around it.

Tom Liotta