I have a question for that of you that are not so new to EJB like I am. We have to write a software with EJB 1.1 (JOnAS 2.2.6), that does the following: - User sees an "Explorer View" (JTree, JTable). In the table he sees all entities (three to five attributes; about 5000 entities). He browses through the entities (including sorting by click on header) and when he found the entity he wants, he double clicks it. An editing windows opens, he changes some data or starts some complex server-side processing of that data. - Server processing is done with SessionBeans on JOnAS 2.2.6 server. Performance of server processing is OK. The problem is: We want to load only that entities that are seen (this is provided by a descendent of AbstractTableModel). This is quite slow, because of many ...getXXX that run outside of transaction boundaries (for the Explorer can be open for hours, the transaction would run in a timeout). So we have to explicitely begin and close transactions every time an entity is wanted by JTable. But this is some kind of slow, because of the many begins and commits. Sure we know that we can do this with direct JDBC from the session bean, but this is no very nice (object-oriented) solution. So our question is: How did YOU solve this? Did you all go over JDBC, or is there some guy that has the ACME solution to this? Thank you Markus
