Aids for program interchange.

The concept of program interchange has been the fundamental idea behind the creation of Software indexes such as the 10,000 strong NCC National Computer Program Index, and hire and buy schemes in the USA. It may be that the realisation of the difficulties involved, combined with the US Navy's decision to use only USASI Cobol and Fortran inspired the lecture. One would have thought that the subject would have been of sufficient interest to attract more than the few that attended. (Editor).

The difficulties in the way of simple program interchange are those of obtaining common standards. The introduction of high level language enabled programs to be rewritten each time they were needed for relatively simple problems. Now problems have become so complex that it is expensive to continue to throw them away. Further sub-divisions of standard languages into dialects prevents successful interchange. For example two Fortran standards exist

X3 9--1966 Fortran which approximates to Fortran TV and X3-io--1966 Basic Fortran which is equivalent to Fortran II.

USASI Fortran has developed from ASA Fortran and has been specifically devised to assist program interchange.

Problems in implementing USASI.

Writers in USASI do not always take into account the limitations placed upon the programmer.

e.g. An apparent Fortran program may be read

EQUIVALENCE (I.J)

I=1

(A(J) = 1.235

but USASI does not allow J as a subscript, therefore this is not USASI

Therefore a program may look USASI but since it is not USASI it is not transferable.

Note: NCC have produced a manual for USASI Fortran which is in print at the moment. The most important part of this manual is the section on the semantic restrictions on USASI.

For example

Function F(K)
      K = K 1
      F = K
  RETURN
   END

But J = I + F(I) is unacceptable to USASI as thus if I = I it may compile as 3 or 4.

Perhaps the most difficult problem as far as the user is concerned is the fact that it is essential that for USASI Fortran to work, compatibility is required between USASI compilers, and some sort of testing procedure is necessary in order to ensure this occurs.

This testing procedure may be

* Syntax check

* Fortran compile and execute.

A cheaper method may be to put the test program through a USASI Syntax checker then test it on several compilers.

Any program interchange must ignore the length of word in the machine and adhere to the rules. System features should not be used that cannot be replaced in pure Fortran.

For useful interchange it is essential that documentation be provided, both user documentation, and language documentation. A set of aids is being provided by NCC that will automatically supply this data.

It includes (1) A Tidy program that renumbers all of the program statements in ascending order--for example if 53 IF@ 54,27,298 is presented then it is certain that 27 precedes 53, 54 follows immediately and 298 is a numerical sequence much further on.

(2) A Flow Charting program which produces empty boxes in order that me originator can put relevant information in the flow diagram in significant terms.

(3) A Dictionary program provides a listing of all names in the program, where they are and what they mean.

(4) A Syntax check is provided.

USASI Cobol is also available as an interchange language. A questionnaire circulated in the UK to ascertain the degree of interest in this language was a waste of time. However in the USA a compiler has been produced which translates any form of Cobol to USASI Cobol, and it is expected that by Feb. 1970 USASI Cobol will be available on every US commercial machine.

Fundamentally however the conclusion that one must draw is that programs must be written with interchange in mind at the beginning, not as an afterthought at the end.

The original text from Vol. No.l 1969

M. B. Wood NCC