Galler: I am concerned that manufacturers use this difficulty as an excuse for absolving themselves of any responsibility for the system after any change the user makes.
Naur: If you want to modify a system, you had better choose a manufacturer and a system that allow this. With most products the standard case is that a warranty is voided if the customer fools about with the product.
Babcock: Often there is no choice for such a system open to the user.
Bemer: This shows the need to improve the means of specifying program interfaces.
Berghuis: A paper in the Communications of the ACM in 1963 or 1964 gives the manufacturer’s viewpoint. This paper, written by the ECMA, the European Computer Manufacturer Association, states that no manufacturer will take responsibility for any user modifications.
6.2. REPLICATION, DISTRIBUTION AND MAINTENANCE
6.2.1. REPLICATION
As discussed in Section 3.3, the replication of multiple copies of a software system is the phase of software manufacture which corresponds to the production phase in other areas of engineering. It is accomplished by simple copying operations, and constitutes only a minute fraction of the cost of software manufacture.
Randell: If software replication costs were commensurate with hardware replication costs, there would be a great incentive for manufacturers to improve the quality of initial software releases.
The main contribution on the subject of software replication dealt with the problems associated with mass production of copies of software.
Enlart: (from Program distribution and maintenance)
»We do not consider in this paper the sophisticated headache of generalized operating system testing. We assume that the development programmers did a good job and tested their product carefully by means of sample problems, selected potential users’ cooperation, bench mark application, etc The development programmers will issue an information medium loaded with the programming system, together with its literature. The literature can be mass produced and stored to meet the forecasted requirements of the potential users. Unfortunately, mass production of information media is not feasible to date, for lack of standardization in machine configurations and input devices.
Another reason is the very low efficiency of the available DP material to mass produce information media. The performances of the fastest input-output devices available now are limited for technological and physical considerations, and, in spite of their high efficiency in terms of data processing, they are definitely not mass-production tools. The last point is the number of interfaces between program authors and users which results in the number of times a master should be duplicated to supply the Program Distribution Centers with their own master copies, enabling them to disseminate copies of the program.
Basically, every bit of information is vital, hence the requirement for the 100 percent checking of every copy. As performance checks cannot be resumed after every copy operation, it is necessary to provide the successive functions which should duplicate a program with tools and means of control, to achieve the highest ratio of reliability.
This quality control problem is worsened by the sensitivity of information media to a variety of mishandling, in or outside of the machine room: there is no ‘acceptable percentage of variations’ or ‘plus or minus acceptable tolerance’ in software: a bit of information recorded on a tape is true information or false information.
If a bit is false, it will spoil the whole product, and, if reported as a bug to the authors, it will cause bewilderment and useless attempts at a corrective solution. In any case, it will ruin confidence (if any) in relations between both parties.