Request for comment : Transaction Sets

[Mike.Sheen]

During the JSP 2008 conference, I talked about the concept of Transaction Components.

This concept has had a slight name change, is now called Transaction Sets, as we think it better suits the function.

Attached is a design document for the Transaction Sets concept we are building for version 7.

Please, download the document (it's a Word 2007 document), and leave any feedback - suggestions, criticisms (or praise!) here.

[David.Simmonds]

Mike, So you plan to have one big table called TR_Set which then refers to 4 other tables - one each for each transaction type. Won't this cause performance issues in a larger multi - user system. Every transaction being written to the database will have the same starting point...
David

[Scott.Pearce]

No, we do not believe that performance will be an issue in this regard. Of course, we will be running benchmarks to be sure.

[Mike.Sheen]

Mike, So you plan to have one big table called TR_Set which then refers to 4 other tables - one each for each transaction type. Won't this cause performance issues in a larger multi - user system. Every transaction being written to the database will have the same starting point...
David

On the contrary, I think it will mean improved performance.

Currently our reconciliation and audit reports need to trawl through the source documents to reconcile the GL against the sub-ledgers (inventory, debtors and creditors).

Using the Transaction Set concept, we - as you put it - have a single starting point. This way reconciliation and auditing is a lot simpler. Isolating discrepancies should be as simple as comparing the GL entries against the sub-ledger entries for a given period, then refining down to a particular transaction set or sets.

Of course, the main purpose of Transaction Sets is to eliminate the cause of any such discrepancies - the software should encapsulate all related transactions together, so it should not be possible for imbalances between the GL and the sub-ledgers.

[David.Simmonds]

Do you think you will get the performance from the simpler table structure with less querying of other tables, so does that mean that the tables will have more data (more non normalised tables) in them, will JIWA 7 be bigger?

[Mike.Sheen]

so does that mean that the tables will have more data (more non normalised tables) in them, will JIWA 7 be bigger?

No - no more data than currently.