Application style
Your FEPI application can have one of three conversational styles.
One-out one-in conversational
One transaction performs the complete conversation with the back-end application in a single send and receive operation. This is the simplest style, if the required data can be obtained from the back-end application in this way. The transaction can be reduced to a single FEPI CONVERSE command using a temporary conversation.
By freeing the connection between transmissions, the capacity of the connection is increased. However, this style only works where no setup is needed to run the back-end transaction and it does not depend on any prior communication. This is because, unless you have a very strict pool regime, you cannot generally guarantee which simulated terminal FEPI will use–it may not be the same one as in a previous conversation–or that you were the last user of the terminal. Further, if you receive unexpected results from the back-end transaction, you may not be able to recover. Therefore, you should only use this style where it does not matter if the back-end transaction runs or not, for example, for a simple inquiry. A one-out one-in conversational program is unlikely to be suitable for accessing CICS® transactions or IMS™ conversational transactions.
Conversational
One transaction performs the complete conversation with the back-end application using multiple send and receive operations and waiting for the inbound data to arrive. This style is used for a back-end application that requires several transmissions or complex setup. This style is simple, and if the network performance is good, the time spent waiting for inbound data may not be a problem.
Pseudoconversational
One transaction sends data to the back-end application, identifies another transaction that is to be started when the inbound data arrives, and ends. When inbound data arrives, FEPI starts the specified transaction which then receives the data. A typical technique is to have a transaction that, when started to receive inbound data, receives the data, sends the next piece of outbound data, issues FEPI START to start itself, and then ends.
The pseudoconversational style (use of FEPI START commands) results in significant CPU overheads in the front-end region. Further, since the use of FEPI START generates additional flows to and from the real terminal, response times are also significantly increased. As a consequence, FEPI START should be used sparingly when, for example. the receipt of the data from the back-end application takes a long time.
See the FEPI sample program: Screen image SEND and START and the FEPI sample program: Screen image RECEIVE and EXTRACT FIELD.