Despite Baker's failure to include them with his rather manufacturing-centric views, the relational technology brought forth many well-acknowledged benefits to the data-processing world at large. Clearly, relational database systems are being used today by a far wider audience of users than would have ever been possible with CODASYL and hierarchical database systems. And, as Ted Codd pointed out in his Turing lecture, considering that the primary motivation for them has been to improve the productivity of application development through data independence and dynamic administration of databases, relational data systems have performed this job very well.
Interestingly, had relational database systems not been popularized, CODASYL and hierarchical database systems would, today, face similar problems to those which Baker attempts to attribute exclusively to the relational realm. Fundamental limitations in data types, as well as the restricted network modeling, make it equally difficult to use CODASYL and hierarchical systems for such applications as multimedia, engineering, genetics, etc. Here, Baker is quite correct in his assessment that object-oriented database technology is the key to delivering support for higher levels of automation and organizational improvement; however, he may be surprised (mirthful grin here) at the way in which these systems evolve.
Today's challenge is to develop database systems which carry forward both the equivalent, or better, modeling power of CODASYL and network database systems and the data independence and declarative query language features of relational database systems. This seemingly oxymoronic task is actually quite possible, as recently demonstrated by the commercial availability of UniSQL, Inc.'s (Austin, Tex.) unified relational and object-oriented database system, UniSQL/X, the UniSQL/X database system supports the SQL/X query language which has been developed by extending ANSI Structured Query Language (DDL, DML & Query, and DCL) with natural support for classes, objects, arbitrary data types, methods, complex nested data structures, and multiple inheritance. To a large extent, the UniSQL/X database system solves the very problems upon which Baker laments.
The core work which led to development of the SQL/X query language, and subsequently to development of the UniSQL/X database system, was presented in a paper, "Querying Object Oriented Databases" by Michael Kifer, Won Kim, and Yehoshua Sagiv, at the SIGMOD/PODS '92 conference that was held in San Diego, Calif.
Albert D'Andrea UniSQL, Inc. Austin, TX
Baker's letter offers a vivid example of the frustration inflicted upon technical experts by salespeople and bandwagon-chasing managers, but adds little new insight on the technical issues. In particular, Baker's main point on the unsuitability of relational databases to support bill-of-materials applications--since they cannot express transitive closures--has been known for a long time (actually elucidated by the very "database theoreticians" maligned in his letter [1, 2, 3]).
Also, any disagreement on how well the relational approach has served the majority of database applications in the past should not make us lose sight of the main message contained in the October 1991 article: traditional database systems (read current SQL) are inadequate to support an emerging new wave of database applications for which new solutions are needed. While in the database field, there is not yet a consensus on the best approach to be pursued (e.g., an OO approach or a synthesis of OO and deductive systems), there is a shared hope that whatever new technology is adopted it will enjoy the same degree of success experienced by the relational technology, which, in the past 20 years, has seen a tremendous technical and commercial growth, and also promoted progress in related technical areas, such as distributed databases and paralle machines.
Carlo Zaniolo University of California Los Angeles, CA
Baker advanced arguments that the classical harmony of pointer-based database systems will be rediscovered. I have experience of a sort similar to his, with a good track record, but I welcomed the advent of fully developed relational database systems with open arms. They appeared to help me with many of my design problems, rather than require me to design around their many restrictions. I was premature, of course, because most relational database management system implementations have gone a step too far toward modernness, and are not as mature as one would hope.
The problem is not with the relational model, but with the implementations of it. They are always written in high-level languages, and often use very poor techniques for access to the dictionary underlying most decent relational systems.
Very few relational database management systems acutally requre the use of a unique primary key for every table. This single departure from E.F. Codd's original theory promotes the desing of inefficient database. The relational database model will inevitably require more I/O activity per request than pointer-based models. This, however, is the price of flexibility. To achieve a similar amount of flexibility with the average pointer-based system, one ends up incurring I/O levels near those of a relational system, and will never achieve equal flexibility. Flexibility is the ability to change the sturcture of the database rapidly, and to respond to queries that had not been thought of at the time the database was designed.
The relational model is not inherently bad. So far, however, there don't seem to be many implementations of it that are as efficient as some of the traditional database management systems. The reality is that today's databases are relational.
Patrick J. Brennan Norfolk, VA
Baker's letter misses an important point. Vaughan Pratt long ago pointed out the existence of a "competence/performance dichotomy." There is a trade-off between the competence of a system--the things the system is capable of doing--and the performance of a system--how fast it is able to do those things of which it is capable.
Sometimes, it makes sense to trade competence for performance; that is the secret of RISC architecture, for example. Other times, it makes sense to trade performance for competence, and that trade turned out to be a big win in the case of relational database systems. One added competence of these systems is a raise in the level of abstraction, which affects both how we program (a few lines of SQL substitute for many lines of conventional code) and who may program (relational query languages are accessible to people who would have trouble programming in C++, for example). A second added competence is the ability to support ad-hoc queries--the ability to respond reasonably well to questions that were not predicted by the designers and therefore could not be "wired in" with pointers. There are many other added competences, such as the ability to restructure data easily.
Interestingly, the clean abstractions of the relational model can lead to performance enhancement; for example it facilitates the use of scalable, shared-nothing parallel architectures for databases. There is also adequate benchmarking to show that current relational systems fare quite well compared with the older systems, although issues other than the model have impact in these comparisons.
Baker, in his criticism of the relational model seems to focus on performance and dismiss the desire of the typical customer for the added capabilities of relational systems. These systems were not successful because of any conspiracy or notational fraud, as the letter in question seems to suggest, but because people vote with their feet and with their pocketbooks. The applications that did not need the added competence stayed with the older generation of systems, while the many applications that required some or all of the additional capabilities made the relational model the profound success it has become.
Baker touts object-oriented systems in his letter. For applications needing competence different from that of relational systems, OO systems may indeed prove a big performance booster. However, history has made clear the need for the competences of very high-level programming and ad-hoc queries in at least some applications. Thus, there is also a need for preservation and enhancement of the relational model's capabilities; the latter approach is being developed in "logical," "rule-based," or "deductive" database systems. In fact, possibly motivated by the evident need for the competence of relational systems, many of today's OO system vendors are moving to provide declarative query capability. We may yet see a synthesis of these paradigms leading to systems with better performance than early pointer-based systems and better competence than current relational systems.
Francois Bancilhon O2 Technology
Philip A. Bernstein Digital Equipment Corp.
Michael Carey Univ. of Wisconsin
David DeWitt Univ. of Wisconsin
Ronald Fagin IBM Almaden Research Center
Hector Garcia-Molina Stanford Univ.
James N. Gray Digital Equipment Corp.
David Lomet Digital Equipment Corp.
Avi Silberschatz Univ. of Texas
Jeffrey D. Ullman Stanford Univ.
Gio Wiederhold Stanford/DARPA
Maria Zemankova NSF
We were pleased to see the May, 1992 Special Issue on Computers and People with Disabilities. The topic is timely and important, and we are very happy that Communications chose to look carefully at it. There were several things in the articles that one or the other of us didn't know before, and we have no major criticism of most of the technical information. However, we have some other criticism and some comments.
We will first comment on Carl Brown's article. His descriptions are basically good, and the topic was covered reasonably well. The product reference list on p. 40, and the list of organizations on p. 42 are also good. We have found that this information is not readily available to the person who is starting out to find out about adaptive aids for visually impaired persons, and we have compiled similar lists , which includes most of the products and organizations on Brown's lists, plus a few others.
There is one product in particular that we think should be added to Brown's lists, the pocket braille notebook computer. This type of device is exemplified by the Braille 'n' Speak (Blazie Engineering, 3660 Mill Green Rd., Street MD 21154, 301-879-4944). This "notebook" has 200K of memory, can interface with almost any other computer and/or peripheral, can fit in a coat pocket, and sells for under $100. It has no monitor; it uses an adjustable speech synthesizer. It does not have a full keyboard; it has 6 keys plus a spacebar. It has software to convert between ASCII and Grade 2 braille. Of course, one needs to be familiar with braille to use it.
With regard to speech synthesizers, we would like to add a clarification to the statement on p. 38. Brown says,"... speech synthesizers are available from many sources and range in speech output quality from barely understandable robotic utterances to well-modulated voices with all the inflection, nuance and color of human speech." This is technically correct, but if access to a computer does not depend on a speech synthesizer, the reader might assume, as the author seems to imply, that more human-sounding speech is better. However, the person who depends on a speech synthesizer knows that "understandability" is not necessarily related to how "human-sounding" the speech is. In fact, a speech synthesizer that can be speeded up may be far more useful to a blind person than one that has the "inflection, nuance and color of human speech." Someone who is used to a particular synthesizer may understand it very well at a rate too fast to be intelligible to another person. This speed is very important, since this is usually the limiting factor on the rate at which a visually impaired person can use a computer. Why worry about whether the speech synthesizer sounds nice to a sighted person? Concentrate on what is important to the user.
Secondly, we would like to comment on a topic that was not mentioned regarding Graphical User Interfaces. At present, GUIs eliminate the functional use of screen reader systems. Current screen reader devices capture ASCII text from a memory buffer in the video adapter hardware. That guarantees a standard location from which to get ASCII text to be converted to speech. In a GUI, such as Microsoft Windows, the video memory contains a pixel-based graphical image. To get an ASCII text it then becomes necessary to find it in another location, or to add a processing layer to handle character recognition. In either case, it makes existing text mode screen readers obsolete. That is a nontrivial requirement both in terms of reliability and speed. Text mode interfaces are currently essential to users dependent on screen readers. We have already heard of a visually impaired data entry clerk who was furloughed when the company decided to switch to GUI-based software. This rendered the previously used screen reader system unusable, thereby preventing the person from doing the job.
Finally, we object strenuously to the quote that Brown uses in his article, which is also emphasized by the headline introducing the special section and partially reproduced on the cover. Its use as a headline indicates the significance that the editors, as well as Brown, placed on it. That quote reads:
"Discrimination against the
handicapped was perceived by
Congress to be most often the
product not of invidious animus,
but rather of thoughtlessness
and indifference--of benign
neglect.... For example,
elimination of architectural
barriers was one of the central
aims of the Act, yet such barriers
were clearly not erected
with the aim or intent of excluding
This editorial error is a classic one made by many nondisabled people, and that error is to lump all disabilities into a single category; in this case, those who suffer from architectural barriers. This emphasis is not supported by the contents of the special section. The section includes one article primarily about adaptations for visually impaired persons, persons with orthopedic disabilities that limit typing ability, and persons with hearing impairment; three articles about various types of adaptive aids for people with speech disabilities; an article about aids for people with some other types of physical disabilities; and an article about other adaptive aids for persons with hearing impairment. None of these disabilities necessarily has anything to do with "architectural barriers."
However, let us assume for a moment that the editorial intent was not to concentrate on one type of disability; the intent was to concentrate on the concept of "benign neglect." While it may be appropriate to indicate that this view was held as recently as 1985, the recognition by Congress of the fallacy of this view could be demonstrated by quoting from the preamble to the Americans with Disabilities Act of 1990 (ADA), Sec. 2, Findings and Purposes, which describes the problems faced every day by people with disabilities, stated in nine different ways. Quoting only four of them:
The Congress find that ...
(5) individuals with disabilities continually encounter various forms of discrimination, including outright intentional exclusion, the discriminatory effects of architectural, transportation, and communication barriers, overprotective rules and policies, failure to make modifications to existing facilities and practices, exclusionary qualification standards and criteria, segregation, and relegation to lesser services, programs, activities, benefits, jobs, or other opportunities;
(6) census data, national polls, and other studies have documented that people with disabilities, as a group, occupy an inferior status in our society, and are severely disadvantaged socially, vocationally, economically, and educationally;
(7) individuals with disabilities are a discrete and insular minority who have been faced with restrictions and limitations, subjected to a history of purposeful unequal treatment, and relegated to a position of political powerlessness in our society, based on characteristics that are beyond the control of such individuals and resulting from stereotypic assumptions not truly indicative of the individual ability of such individuals to participate in, and contribute to, society;
(9) the continuing existence of unfair and unnecessary discrimination and prejudice denies people with disabilities the opportunity to compete on an equal basis and to pursue those opportunities for which our free society is justifiably famous, and costs the United States billions of dollars in unnecessary expenses resulting from dependency and nonproductivity. [emphases ours]
It is probably true that buildings were once designed without thinking about access for wheelchair users. However, this has nothing to do with the problems facing people who are visually impaired, hearing impaired, speech impaired, or have other physical disabilities. It may even be true that there is such a thing as "benign neglect" of people with disabilities; however, it is far overshadowed by intentional and purposeful discrimination, and no one who has a disability or who has lived with and/or cared about someone with a disability would claim otherwise.
To your credit, you cite an ACM Membership Survey which deals with many types of disabilities, concentrating on viewers' ability to read a computer monitor. This is probably the primary limitation for people with disabilities who work with computers, and the emphasis is a logical one. The survey then looks at other types of disabilities and the accommodation requirements for them, including architectural barriers. This shows that someone in the ACM understands that this type of barrier is one of many problems, and that disabilities come in a variety of forms.
In conclusion, thank you for the issue addressing disability concerns, some of the technological adaptations that are available, and research that is currently being done. Perhaps next you will also consider the problems of social acceptance and real job opportunity that we face.
Charles D. Sigwart, Gretchen L. Van Meer Northern Illinois University Dekalb, IL
I was delighted to read Carl Brown's article, "Assistive Technology Computers and Persons with Disabilities," (May 1992, p. 36). Brown observed that "developing screen readers for graphical Windows environments for MS/DOS computers remains and unanswered challenge." I assume he is referring to the Microsoft Windows environment. While I do not know of a specific system for that environment, I believe I can claim that we have not only addressed but in fact answered the challenge.
A group led by James W. Thatcher at the IBM T.J. Watson Research Center has spent the last few years developing a screen reading system for the OS/2 Presentation Manager (PM). It has been demonstrated at several conventions and conferences since early 1991. The current prototype running under OS/2 2.0 provides full access not only to OS/2 PM applications but also to Microsoft Windows applications running within OS/2 environment.
For further information on this project, you can contact Frank Dipalermo at IBM's Special Needs Systems group in Boca Raton, Flor. His phone number is (407) 982-0006.
David Jameson Thomas J. Watson Research Center Yorktown Heights, NY
(1.) Aho, A. and Ullman, J. Universality of data retrieval languages. In Proceedings of the Sixth ACM SIGACT-SIGMOD-SIGART Symposium on Principles of Database Systems, 1979.
(2.) Chandra, A. and Harel, D. Structure and complexity of relational queries. JCSS 25, 1 (1982), 99-128.
(3.) Fagin, A. Monadic generalized spectra. Zeitschr. f. math. Logik und Grundlagen d. Math. 21 (1975), 89-96.
(4.) Van Meer, G.L. and Sigwart, D. The Implications of the Americans with Disabilities Act of 1990. In Proceedings of the Midwest Computer Conference. (Hammond IN, Mar. 27), 1992.
|Printer friendly Cite/link Email Feedback|
|Author:||D'Andrea, Albert; Zaniolo, Carlo; Brennan, Patrick J.; Bancihon, Francois; Bernstein, Philip A.; Car|
|Publication:||Communications of the ACM|
|Date:||Aug 1, 1992|
|Previous Article:||Editorial; Communications of the ACM: here we are, where we are going.|
|Next Article:||A performance comparison of four supercomputers.|