Review of     "Development of Executive Routines, Both Hardware and Software"
              by Albert B. Tonik; (AFIPS Conference Proceedings 31, pp. 395-408,
                                  Thompson Book Company, Washington, D.C.)


      Innumerable papers by Russian authors using probabilistic models for physical
systems are prefaced by a comment explaining that the results obtained must follow
from dialectic materialism, and therefore the same results could in principle be
derived deterministically as well as statistically, save for mathematical con-
venience.  A similarly political motivation may unconsiously underly many papers
in the computer field which explain that ever more elaborate executive programs
are motivated exclusively by a desire to increase hardware efficiency and improve
system economy.

      Thus this paper, an otherwise excellent review of the array of problems
encountered in organizing an executive program, weakens its semi-historical develop-
ment by presenting each elaboration of executive responsibility as motivated only
by improved hardware utilization.  In the most serious example of the distortion
caused by this viewpoint, the paper suggests that the motivation for paging as
a scheme for allocating primary memory is to sandwich additional programs into
memory in the holes left between other programs.  Whether or not paging increases
or decreases the number of words of primary memory which can be utilized probably
depends on the statistics of the programs involved.  The more serious issues (and
the ones which raise the most emotional arguments) are the abilities of a paged
machine to present the programmer with an interface which does not change every
time a new box of memory is added; and to relieve the programmer of explicit
overlay organization.  The author mentions these latter issues as though they were
virtually side effects of the effort to increase hardware utilization.

      With this single significant complaint aside, this paper is a welcome relief
from the long series of "here is how we implemented a multi_____ system on the
____ computer".  The author has taken the trouble to identify many of the sources
of difficulty and complexity in an executive program.  This is one of the few papers 
one can point to which recognizes the importance of error handling and program
snapshot/restart procedures in their contribution to the complexity of executive
programs.

      A discussion of solutions as comprehensive as the author's discussion of
problems would require at least a book.  The author has taken what appeared to
be an easy shortening scheme--he indicates one or two techniques of attacking
each problem.  Unfortunately, he becomes unintentionally dogmatic when he presents
the technique he is most familiar with and fails to indicate that there may
exist several other techniques to solve the same problem.  This comment applies,
for example, to his discussions of priority schemes, and of error handling, and
of interrupt handling.

      Overall, the paper provides a lot of food for thought, and is well worth
exploring before implementing another operating system.  A comprehensive
bibliography of papers on operating systems provides valuable reference.



                                                       J.H. Saltzer
                                                       Waban, Mass.


5/13/68