For the Record, Please
Self-Assessment Procedure XVII. Tut, tut, ACM! I am not merely a "famous person" and Turing award winner (Communications, Oct. 1987, pp. 886-891). My highest distinction is that I am a past ACM publication editor. I was an associate editor of journal from 1967-1969, and I edited the August 1966 special issue of Communications devoted to papers on symbolic and algebraic manipulation, the first issue that exceeded 100 pages of editorial material. Assess thyself!
Robert W. Floyd, Dept. of Computer Science, Stanford University, Stanford, CA 94305
The procedure possibly asks the self-assessor to distinguish between "publication editor" and Publication Editor--Ed.
ETA Still Off-Course
I have just read the two "Profiles in Computing" in your October 1987 issue (pp. 816-23) and was impressed with the fact that Donald Knuth and Brian Reid both mentioned the quality of the programming in the airline industry. I agree that the reservation systems are impressively reliable, but find the real time reporting of airplanes in flight, or waiting to take off, woefully inadequate.
In these days of airport congestion and delayed maintenance it is important that a party meeting a flight at its destination be able to obtain accurate predictions of its arrival time. Although the pilots can make these predictions to within five minutes once they have taken off, the reports from the airline reservation and information system at the destination city are often off by an hour. If the same emphasis on reliability were put on the flight reporting systems, a great deal of the frustration from delayed flights would be removed.
Nelson L. Max, Lawrence Livermore National Laboratory, University of California, P.O. Box 808, Livermore, CA 94550
I read with some amusement the article entitled "Cobol on a PC: A New Perspective on a Language and its Performance" in the February 1987 edition of Communications, pp. 142-154. I am surprised that Paul Jalics did not consider any of the PC versions of MUMPS as alternatives to Cobol.
In an article by Thomas Munnecke from the 1980 National Computer Conference entitled "A Linguistic Comparison of MUMPS and Cobol," it takes 8 percent of the code to write a payroll application in MUMPS versus the same application in Cobol (300 lines of code versus 3600 lines of code). In programmer's hours, the MUMPS system saved five to eight man months. Not only that, but the MUMPS system ran at 50 percent efficiency on a minicomputer versus Cobol on a mainframe.
In the January 9, 1984 Computerworld article entitled "A Case for MUMPS," Casimiro Alonso found equally impressive results. Comparing MUMPS and Cobol on a VAX, database creation was 13 minutes versus two hours, data base query was 1 minute 37 seconds versus 2 minutes 35 seconds, data base disk occupation was 5.4K versus 21.7K, and lines of code needed were 183 versus 1050.
I am surprised you printed Jalics' article. What do we get next month, "Transistors, the Wave of the Future"? Keep up the scholarly work.
Jeffrey P. Wellmall, Vice President, COMP Mark Inc., 1601 Westheimer, Suite 201, Houston, TX 77006
I commend Jalics and ACM for addressing the topic, Cobol on the PC, that is of immediate practical importance to many of us. He presents some startling information and has used what appears to be real world test data, not classroom problems. Now, some comments and suggestions.
His choice of the Realia compiler did not surprise me because there have been several evaluations published in the slick PC magazines supporting its implementation level and speed. What did surprise me is the statement that the compiler is written in Cobol! I am sure that those of us who write quite a bit of Cobol and also a bit of assembler, C, etc., would love to read an article explaining some of the techniques used by the compiler writers. Many of us could probably write Cobol in Cobol but it sure would not be fast.
The sample code in the article was rather hard to read because data definition statements were not included. Page 152, line 4 contains an obvious error in the PERFORM statement. The examples use subscripting rather than indexing; some comments on the relative performance of the latter would have been helpful. Perhaps an unintended result of the comparison between Cobol 74 and Cobol 85 is that, to me, the Cobol 85 is harder to read. Removing those paragraph names also removed valuable documentation.
I am sorry that Jalics had to labor with both an obsolete operating system (OS/VS1) and ISAM. VSAM would certainly have changed the I/O comparison.
Perhaps the most serious problem with the article that will definitely change all the tabulations of mainframe numeric field performance was misinformation on p. 144, the decimal truncation anomaly. The statement on truncation of fields nine digits or less is just not true. For example, S9(4) COMP fields are truncated, IBM faithfully preserved the decimal definition of Cobol in its implementation of the binary types and the IBM compilers do generate much spurious code to convert binary items to decimal and truncate any digits exceeding the picture definition, However, there is a compile-time switch in all the mainframe compilers that shuts off truncation; its use produces vastly improved performance when extensive manipulation of binary data is performed. Setting this switch will always eliminate a convert to decimal, a compare, and convert back to binary.
Instructions to expand the binary to a full-word and such may also be eliminated. My experience has been that much more CPU time is used to truncate than to do the arithmetic. For OS, the JCL for the compile should include PARM.COBNOTRUNC; for DOS there is a CBL card option. Most IBM shops which I have been involved with set NOTRUNC as a standard compile option.
Overall, I think that the article addressed a real need and I am sure that you will get more correspondence. My hope is that interest by ACM members will help Jalics revisit the subject and publish an update.
Francis J. Tague, 51 NW 17th Street, Homestead, FL 33030
I also read Wellman's letter "with some amusement." My article was not trying to promote Cobol over other programming languages. I am very much aware of its problems as well as its potential. Cobol is simply a reality as a language used for the majority of business applications. Instead, my main focus was on the coming of age of PC Cobol's and the study of specific performance characteristics of Cobol on PC's in relation to characteristics seen on mainframes.
Would MUMPS be a good replacement for Cobol? I am not in a good position to judge, but I sense some frustration on why the world isn't talking more about this possibly better tool. In that vein, I want to remind Wellman that technical achievement is no guarantee for market acceptance; for example IBM, possibly the most potent sponsor, presented us with a very suitable replacement for Cobol in 1965 (PL/I), and PL/I has gone nowhere since then.
1. Line 4 on page 152 should indeed be:
PERFORM GET-A-MAIN UNTIL A-EOF = 1 OR A-FOUND = 1.
2. I have rechecked my facts regarding the decimal truncation anomaly referred to on p. 144. My facts are correct for the compiler I have been using. Furthermore, I have rerun the experiments under OS/VS Cobol Release 2.4 dated July 1, 1982, the latest release as indicated to me by IBM. The results are still identical to my published results.
3. The compiler option NOTRUNC was on (it is the default value as described in the Cobol Programmer's guide SC28-6483-1). I believe in using the default options in such measurements, since I think these are the ones used by the typical programmer. I have also rerun the experiments by setting TRUNC on explicitly. This has the effect of quadrupling the ADD time for fields under 10 decimal digits and makes little difference for fields of 10 or more digits.
4. I agree that whether my Cobol-85 sample is more or less readable than the Cobol-74 one is a matter of opinion. My main point here is that the code looks more like some other programming languages that we think of as more up-to-date like Pascal, Ada, or C.
5. I have measured the performance impact of subscripting versus indexing on a number of machines (see references 1 and 2 in the article), and the results vary widely from machine to machine. On the 370, Cobol's indexing is usually more efficient (by 15% in one test), whereas on the Realia PC Cobol the performance is about the same. Generally speaking, subscripting is nearly or equally as fast as indexing, given that the subscript is a natural data-type for the given machine. For example, a 32-bit binary integer (PIC S9(9) COMP) is the optimal subscript on an IBM370 architecture, and a 16-bit binary integer (PIC S9(4) COMP) is the natural datatype for an IBM PC.
6. I intend to rerun my experiments on an up-to-date MVS system using an up-to-date Cobol compiler on an IBM3081 and an 80386 based PC as soon as the above are available to me, and publish the results.
Paul J. Jalics, Associate Professor, Cleveland State University, Dept. of Computer and Information Science, 2121 Euclid Ave., Cleveland, OH 44115
This letter is in response to a letter from Joan Marie Verba in the July 1987 Forum.
I have been an ACM member off and on (more on than off) since 1970 and have worked in a multiplicity of jobs around the world since 1969. 1 have never been without work involuntarily for more than one month, although I have had longer periods of time when I have voluntarily been out of work. I support a family of four who have traveled with me wherever I have chosen to go to find work. I originally received training in scientific programming (with an emphasis in numerical analysis), but have never held a single job in that category and now doubt my own ability due to lack of experience. But that has not stopped me from obtaining work in the computer field. As a result, I have become somewhat of a specialist in commercial applications so I can work in any non-scientific location without problems.
Her complaint is one I find common with people around the world. It is that she does not have a job in her area and feels she will not be properly qualified when and if one eventually becomes available in her location. She states outright that she does not wish to go where the jobs are available or work in an alternative computer field until jobs in her capacity become available. This somewhat insular attitude is not uncommon. However, what annoys me is while she is using an internationally circulated forum to bewail her woes of not having work in her own locality, she is, apparently, refusing to travel to an area where the work is not only available but sadly lacking in people with the necessary expertise to handle the available work.
A case in point is the country where I am currently employed. New Zealand has better weather than Minnesota. It is small in population, with slightly over 3.5 million. With that population base, it is lacking in people with a high level of technological abilities. Here, there are more jobs available for experienced people in the computer fields than there are people to fill those positions. For instance, the city where I work, Wellington, has a population of about 350,000 and less than lo people with greater than 15 years experience in the computer field. A few months ago I read in a local paper that there were over 6,000 more jobs for qualified computer people than there were people to fill those jobs. But people with the requisite abilities do not come here since they do not wish to leave their own particular locality, where jobs are few and far between.
Although the situation is good for me, I still find it inconceivable and inexcusable that a person truly sincere in wanting a job in their chosen field is unwilling to go where the jobs are available. A person like that will always find that people who are willing to travel to where the jobs are will be the ones to get the choice positions. There may be other ACM members who are having difficulty finding work after having been laid off, but I suspect the majority of these people are the ones who refuse to go where the jobs are but instead want the jobs to come to them. If Verba was willing to travel, then I suspect she would have no problems in getting work and would not have to complain about a lack of work in her field.
Hugh D. Amos, Information Services Division, New Zealand Railways, Private Bag Wellington, New Zealand .
P.S. on page A-21 of the same issue there is an ad for RSVP Services, I wonder if Ms. Verba has tried them.
In the July Forum, Joan Marie Verba of St. Paul complains of lack of employment opportunities. ACM should do far less for its members than for those the members serve. "Jobs for the boys" is not the name of the game,
Verba is not alone in finding prospective employers unimpressed by her ACM membership. I once had a manager, a programmer of several years' experience, who did not even recognize the name, Something is wrong.
Verba does "not wish to work in an area of no interest to me whatsoever," i.e., in financial data processing. This reminds me of a church organist who complained bitterly about his displacement by a friend of the bride at a wedding. The world does not owe us employment as organists and programmers, let alone as church organists and scientific programmers. A capitalist may pay the underemployed full wages (Matthew 20:1-14), but to pay excess sycamore pruners while there is barley to harvest robs consumers.
However, Verba may be mistaken about her own loves. She seems to have a prejudice, common among scientific programmers, against Cobol, together with a perhaps snobbish aversion to DP. I myself worked six years in a scientific environment and nearly two years for a hardware manufacturer before migrating to commerce and administration. I have enjoyed myself about as much as I did in botany and demography. I am happier with Cobol than with Fortran IV. I brought much that was useful to my new work and happily applied it (to the horror of some who had experienced only DP work). Leaving science for DP is not like changing one's spouse. No betrayal is involved.
Work in DP may hurt one's chance of getting back into scientific programming some day. Still, DP will be less harmful than secretarial work.
John A. Wills, Administrative Computing Center, 416 Robbins Bldg., Pasadena City College, 1570 E. Colorado Blvd., Pasadena, CA 91106-9980.
It has been a few months since Joan Verba's letter  appeared in Communications. Most of what Verba described is presumably the result of an economic downturn in the computer industry. There is really nothing ACM can, or should, do about that. ACM is neither a lobby nor a union, and should remain that way.
Verba does mention her suspicion that employers may prefer to hire persons with one to two years experience, again for economic (salary) reasons. Employer preference is the area which I will comment on.
I have encountered some very disturbing employment criteria in the seven years since I graduated with a B.S. in CS. A common complaint is that I am "overqualified"--not in experience, but in breadth and depth of knowledge. Of course, this is a ridiculous excuse for not hiring someone; it strikes me as one possible instantiation of Will Roger's statement "Everyone is ignorant; just on different subjects!" (simply replace 'ignorant' with 'overqualified').
With regard to experience, the responses are even more amazing. You must have at least Y years of experience on machine M using software S, or you are automatically disqualified. If your experience is any other tuple, say (Y, N, S), the response is the same. Even some firms who recruit CS graduates on university campuses will refuse applicants on "lack of experience." The advice given is to get an entry-level job with a firm that accepts people without experience, but that is close to the null set. The Gospel of Employment has rewritten John 3:3 : "Verily, verily, I say unto thee, except a man be born with experience, he cannot see employment."
Most employers I have encountered consider a CS degree an asset, but a disturbing number count it as having no value or as a negative attribute. In their eyes, the problem solving methods learned in CS simply do not apply to their "real world." To them, a degree is a formality, much like being born; it is of no significance. One employer told me if he had to choose between an applicant with ten years experience but known to be incompetent as a programmer and one with two years' experience but known to be competent, he would choose the former because experience was of paramount importance.
While my experiences must be regarded as anecdotal, I do wish someone would conduct a formal investigation, modeled after the Taulbee Survey, into the nonacademic employment of computer professionals. The recent Taulbee Survey  indicates there will continue to be a shortage of Ph.D.s for the near future. What is the situation for graduates with B.S. or M.S. degrees? Also, trade school graduates seem to be in demand. How does their training (one hesitates to call it education) and earnings compare with Computer Science graduates? Surely this information would be of value to CS departments and students as well as employers.
If ACM wants to do many of its members a favor and at the same time induce spinoff benefits for the employers of its members (such as more robust code, higher productivity, etc.), it might do well to educate some employers.
James J. Fullerton, Institute of Logopedics, Research Division, 2400 Jardine Dr., Wichita, KS 67219
1. Gries, D. The 1985-1986 Taulbee Survey. Commun., ACM 30, 8 (Aug. 1987), 688-694.
2. John 3:3. The New Scofield Reference Bible, Authorized King James Version. Oxford University Press, 1967.
3. Verba, J. Employment in the Computer Field. Letter in Commun. ACM 30, 7(July. 1987), 585.
I am writing in response to a Forum letter from Joan Marie Verba. In that letter, Verba stated she is out of work and asked if the ACM is doing anything for its members in similar situations.
Well, Ms. Verba, the ACM will do anything it possibly can to see that you get employed, provided you
meet the following conditions:
1. You live outside the United States, preferably behind the
2. You do not belong to ACM.3. You are out of work because of a political dispute between yourself and the rulers of your country.
Ms. Verba, the ACM is prohibited from doing anything for its members because of its tax status, or some such gobbledygook. In fact, it can be argued that the ACM has done more to ameliorate the conditions of nonmembers living overseas than for its own members within the United States.
Many ACM members lost their consulting businesses on January 1, 1987 (Kristallnacht?) because of Section 1706 of the so-called Tax Reform Act. By virtue of being computerists, ACM members do not have the same rights to form a business like jugglers, clowns, lawyers and physicians. Did the ACM raise its voice like the IEEE or the National Society of Professional Engineers did in protest of such an unfair and discriminatory law?
Well, nobody ever accused the ACM Board of unpredictability, because if legislation is passed that does not affect academics and nonmembers, you can be sure the ACM Board will sit on its corporate fanny.
Naturally, I protested this inaction in a phone conversation with President Paul Abrahams. At least his answer was entertaining. Effectively, he told me if you want to belong to an organization that does things for its members, join the IEEE.
I think it has become apparent that the mission of the ACM is to have its working members subsidize those persons whose livelihood depends on the promulgation of their esoteric trivia.
However, for those of us who have to work for a living, for those of us who must meet almost ridiculous daily deadlines, and for those of us outside of academia, I think it best we take President Abrahams advice.
Michael Tenenbaum, 49 Blanche St., Plainview, NY 11803
Michael Tenenbaum's letter is an unfortunate misstatement of the facts in two respects: what ACM does for its members, and the effects of Section 1706 of the Tax Reform Act on computer professionals.
ACM does many things for its members. Primarily, it provides them with technical information through publications, conferences, and educational programs. Beyond that, it defends the health of the computer field and the right of computer professionals to pursue their profession, whether they are in the United States or abroad and whether they are ACM members or not. This right encompasses both issues of employment and more general issues of scientific freedom. Were a case similar to that of Anatoly Shcharansky to arise in the United States, I hope and expect ACM would act just as vigorously.
ACM's tax status was chosen to reflect the beliefs of our founders about what sort of an organization we should be, not the other way around. ACM is not a professional society in the sense of the American Medical Association or the American Bar Association. Promoting the business interests of our members and lobbying for legislation of economic benefit to them are not part of our charter. Thus, I suggested to Tenenbaum that if this is what he wants, he should turn elsewhere than to ACM.
Thanks to the lobbying efforts of individual computer professionals and groups such as the Independent Computer Consultants Association, the Internal Revenue Service has "clarified" Section 1706 so that it now does not affect independent consultants who contract directly with their clients. Those who are still affected are computer consultants who do contract work arranged through third-party brokers, an arrangement known as job shopping. The broker bills the client and then pays the consultant an hourly rate, which includes the broker's commission. Section 1706 requires the broker to treat the consultant as an employee and deduct payroll taxes accordingly. It also prohibits the consultant from treating himself or herself as a self-employed independent contractor for tax purposes.
Most of these consultants work alongside regular employees of the client, and what the consultants do is almost indistinguishable from what the regular employees do. Section 1706 may deprive these consultants of tax benefits they previously enjoyed and in some circumstances force them to accept less favorable conditions of employment. I doubt it will put any of them out of work, although there are still some real questions of fairness that I will not try to answer here.
It is not correct to characterize the issue as academics versus workers, as Tenenbaum does. The number of ACM members who work as permanent employees is probably far greater than the number who work through third party brokers, and the economic interests of the two groups are not the same.
The issue of Section 1706 was explicitly considered by the ACM Council in February, 1987. The Council concluded that ACM's appropriate role was as a source of information rather than as an advocate, given the nature of the issue and the conflicts about it even within the computing community. I believe most members prefer that ACM remain a scientific and educational society, although I personally think a Question of Importance to settle the issue would be very useful.
Paul W. Abrahams, ACM President
|Printer friendly Cite/link Email Feedback|
|Author:||Floyd, Robert W.; Max, Nelson W.; Wellman, Jeffrey P.; Tague, Francis J.; Jalics, Paul J.; Amos, Hug|
|Publication:||Communications of the ACM|
|Article Type:||letter to the editor|
|Date:||Jan 1, 1988|
|Previous Article:||The state of ACM (President's letter.)|
|Next Article:||Calls for papers.|