Hi there.

I need some input/design suggestion and/or comments regarding 2 projects that I'm undertaking currently. The project duration is 8 weeks starting now.
Some background information of the current technology we are using right now. We are using weblogic 7.0 SP2 and our database is ms sqlserver 2000 and/or mysql. The persistance layer/framework we are using is a home grown JDBC framework. We are using only stateless session beans and our client which is a swing ui access the servers side thru webservices via a session facade. We use also Transfer objects on the client and our own DataObjects on the server side. Current consideration is that we can not upgrade to the current weblogic version at this time and can only use java 1.4 and use existing persistence framework we have. I've considered using Spring/Hibernate but it will involve major refactoring on our existing system, which we can not afford right now. But we can augment or add new framework to handle this 2 projects.

1. The first project is a mass update/creation of orders and order items. In our system the update and creation of order and order items are complicated and perform several validations (e.g. vendor/client is valid; item are valid and have enough to on-hand, etc.) and lookups before we can create/update an order and it's order items. We estimated that process to be around 1.5 - 3.0 secs and typically we will be processing between a few hundreds to about 5000. The max ceiling we put is 10,000.

Now the requirements are:
a. Ensure that the process will be completed and in the event of failure (physical or network or power) and recovery to start where it failed and start processing from that point of failure.
b. The transaction should be atomic on per order basis and not on whole process - meaning we will commit all changes for an order and start a new transaction on the next order. This will ensure that we have a complete order committed.
c. Don't tie the client from this process and when the process is finished give a feedback to the client who invoke the process.

2. The second project is related to the first, is that before they post the process above, the user will look at the set of data and do some editing and then post the data. So this project is on the UI side and would present the user a table/grid view of the data and since data set could be large, we want our swing ui to fetch only a set of rows/data so that it does not take so long to load all the data or run out of memory. A button or by the user scrolling on the scrollbar would fetch the next succeeding rows/data. The problem we have as stated on the first paragraph is these- we are using webservices and stateless session beans and there is no way that the server can recognized who the user is when that user request the next set of row/data set without querying the database again.

So your input and help is very much appreciated and if you know an existing framework that would help me is most welcome.

Have a blessed day.

Jun Victorio