TRANSFER Programming Manual

9 Designing and Writing a Client

069138, Update 1 to 040970 Tandem Computers Incorporated 9–1

Every TRANSFER application must include one or more clients. Clients allow

correspondents to communicate and interact with the TRANSFER delivery system.

From a programming perspective, your client will accept requests from

correspondents and perform the following tasks:

Initiate a TRANSFER session for the correspondent

Translate correspondents’ requests into TRANSFER units of work (UOWs)

Package UOWs into SEND or WRITEREAD commands

Issue the SEND or WRITEREAD to the TRANSFER interactive server (TISERV)

As you design your client, consider the following:

The client divides its attention among the correspondents; therefore, the total

number of correspondents affects the time in which the system can respond to

each.

If a client represents correspondents that are processes, those processes can be

PATHWAY server classes, and possibly agents. If a correspondent is a device, the

process that issues requests to the device can be either a server (possibly an agent)

or a terminal simulator. In any case, a server requires a SCREEN COBOL program

to make requests. Each copy of a SCREEN COBOL program must be assigned

either a physical terminal or a terminal simulator. Many of these applications

might not require PATHWAY.

Any request on behalf of a correspondent must be directed to a server class

running at the node where the correspondent is registered.

This section describes client development steps and includes special considerations for

COBOL and SCREEN COBOL clients. The information in this section also applies to

agents that communicate with TISERV. A sample client is included at the end of the

section.