TRANSFER Programming Manual

Planning Low-Level Implementation
Developing TRANSFER Applications
40970 Tandem Computers Incorporated 8–13
Because TRANSFER groups records into blocks rather than storing each record
individually, these TMF locking limits do not always correspond exactly to the
number of records per item, items per folder, and so on, that are actually
permitted in a particular case. In the worst case, the values as specified by the
TMF locking restrictions apply. In many other cases, the following rules might be
helpful. These rules apply only if the data is added to the objects sequentially.
The amount of data in the item data records added to a single item should
total less than 500,000 bytes. As an example, 6200 80-byte data records per
single item are permissible.
The number of recipients per package, or number of members per distribution
list, should be less than 5000.
The number of items per folder should be less than 500.
These values are overall limits that will allow the TRANSFER delivery system to
process objects without exceeding the TMF locking limits.
If your system manager has not arranged for TMF auditing exactly as
recommended in the Transaction Monitoring Facility (TMF) Management and
Operations Guide, you should note additional considerations in dealing with
TRANSFER files. For example, if the SESSIONS file in the TRANSFER database is
not audited, aborting a transaction that includes the START-SESSION or END-
SESSION UOW might leave extraneous data in the database. This could result in a
WASTEBASKET folder that does not belong to any current session. As another
example, if the ITEM-DATA file is not audited, item data records that no longer
belong to any particular item might remain in the system after the item has been
deleted.
When you use the TRANSFER delivery system to deliver packages over a
multinode network, TMF must be configured at all nodes where the TRANSFER
delivery system is running.
For additional information on TMF transaction processing, refer to the PATHWAY
Application Programmer's Guide.
Planning Low-Level
Implementation
Low-level implementation decisions involve packaging your application in individual
modules (COBOL or SCREEN COBOL programs, FORTRAN subroutines, TAL or
Pascal procedures, or C files) and subroutines, breaking the application up into small
groups of functions. They also involve carefully planning the interfaces between these
modules and subroutines and deciding what information is passed between them.
You can plan for the execution of various accounting procedures, collection of usage
statistics, and logging of particular events and errors. During the test phase of your
development cycle, you might want to log every SEND operation requested by a
SCREEN COBOL program to provide extensive debugging information.
If a large group of programmers is working on the application, plan and establish
coding standards for the group. These standards prevent implementation conflicts,
recoding, and readjustment as the development phase progresses.