Re: open source PostgreSQL not supportable? (Was: Challenging SQL Query Problem. Can you solve it?)

On Fri, 06 Jan 2006 11:20:18 -0800, MBPP wrote:

> I am not UNIX expert but can we write somehow a Oracle data file while
> the instance is up?

Yes, you can.

However, that would almost certainly be detected (immediately if you had block checksums on) as a block corruption. Which almost invariably would cause the DBA to initiate recovery, reversing the manual change.

And you would need to write the change's counterpart into the rollback/undo segment as well as make appropriate entries in the redo - probably by a direct write into the in-memory redo buffer. Otherwise you would have to understand the internal format of the redo log file and it's 3 levels of checksumming.

All this would have to happen before the redo log file was packed up and shipped to the archiving location(s), otherwise it would need to happen consistently in all archive log copies. And that would be a typical window of a few minutes.

Add data-level inline encryption that can even stop the DBA from being aware of the data contents at the SQL as well as file level, and you raise the bar.

All this is possible, but unlikely - sounds like iot would need the Mission Impossible team.

Again - the objective seems to be detecting *whether* the data is reliable, not so much what the data change happens to be. (In other words, has someone attempted to shred documents, or change the facts.) And that capability seems fairly solid,

-- 
Hans Forbrich                           
Canada-wide Oracle training and consulting
mailto: Fuzzy.GreyBeard_at_gmail.com   
*** Top posting [replies] guarantees I won't respond. ***