Hello Tom! Hope your vacation was great. Also, thanks for your response on my prior question.

This is my new question: In our application a user may update/insert/delete row(s) from many tables. If we want to replay the user action we need to trap the user transactions. One other problem is all the users have their own id but backend server which does work for them always logs into db with the same userid. Thus using LogMiner etc very difficult. Also I dont have any say in asking for code change to reflect such information. Mainly if this can be done on database non-intrusively then it is great.

I was thining along following lines
1) create a sequence to be used as transaction number (unless there is such thing as a transaction number accessible in trigger)
2) use a global temporary table.
3) Put triggers on each table such that each checks if a row exists in global temp. table. If one exists then you know some other trigger fired before you and you are either in the middle of transaction or at the end of one. If no row exists then insert sequence number in it and use it in "4" below.
4) Of course after every trigger fire, we will save all table row information and transaction number and sysdate in txn_info table.
5) at the end of transaction the row in global temp will be gone unless it has some use

So I should have a trail table like below

TXN# TABLE_NAME ACTION TXN_DATE OLD_VAL NEW_VAL
1 CUST Delete 09/08/2001 12:12:30 some some
2 CUST_ADDR Delete 09/08/2001 12:12:35 some some
1 CUST_ADDR Update 09/08/2001 12:12:36 some some
2 ACCT_INFO Add 09/08/2001 12:12:36 some some
...

etc

So now if I say
select TXN#, TABLE_NAME, ACTION from tran_info
order by txn#, txn_date

i should get all the transactions.

Do you think this will work or is there any better way? Mind you I have really not developed any complex triggers or such application.

thanks for sea of information you provide.

Sudhir