Table of Contents

December 2001
September 2001
September 1995
March 1995
December 1980
March 1976

What was the last scientific article that you read? Think about it. Not the last article that you paged through or skimmed or sampled the introduction and conclusion, but one in which you actually read most of the pages. Now, how long was that article? Was it a 10-12 conference article, a 20-page journal article, or a 50-page mini-monograph?

Glenn Ellison of the MIT Department of Economics presents compelling evidence (Evolving Standards for Academic Publishing: A q-r Theory) that "In almost all fields papers seem to be longer now than in 1975." Ellison sampled journals in some thirty-two fields and found significant increases in average page length. For example, ACM JACM papers increased in length from 12 pages to 29 pages (though CACM papers actually decreased from 7.6 pages on average to 7.0 pages).

Figure 1 provides more detail for ACM TODS. The top line states the length in pages of the longest article in each yearly volume, the middle line indicates the average length, and the bottom line states the length of the shortest article.

Figure 1: Article length per volume in ACM TODS

Quite frankly, all three trends are disturbing. The average article length has more than doubled, from 19.2 pages in 1976 to 41.9 pages this year. The average article this year was longer than the longest article in 1976. The shortest article this year, at 31 pages, was longer than the average article for the entire first decade of TODS' existence. In five separate years an article of at least 60 pages appeared (one weighed in at a whopping 79 pages).

The total page count per volume was relatively stable over this period, which meant that an increasing average article length was coupled with a decreasing number of articles, as shown in Figure 2. The average issue of TODS in the eighties contained six articles, while in the last seven years the journal has averaged only three articles per quarterly issue, or a paper a month.

Figure 2: Number of articles per volume in ACM TODS

The result is that readers are confronted with less diverse and more ponderous papers in each issue, of concern to the Editorial Board. (A related phenomenon, also of concern, is that referees are confronted with longer manuscripts, which must lengthen reviewing time.) Hence, the question that opens this essay.

The cause of these trends is clear: in an effort to achieve the "significance of contribution" required of acceptance, authors are adding more detail, more theorems, more performance studies, and more discussion to their submissions. (I sheepishly admit that one of the data points in the top line in Figure 1 is a paper of mine.)

It is less obvious that longer papers are better papers. The top five most-cited TODS articles, by Chen, Smith&Smith, Codd, Shipman, and Stonebraker/Wong/Kreps/Held, average a hair over 30 pages each (again, less than the shortest article this year).

Over-simplifying, there are five basic types of papers, appropriate for different venues. There are highly innovative, initial-cut papers that have neat ideas that haven't yet been fleshed out; in fact, it may turn out that the idea doesn't work. Workshops are an ideal venue for such papers. The next step is to pursue the idea further, prove some things or test the concepts out in an implementation, realizing a paper appropriate for a conference. An idea that has merit can be elaborated and detailed evaluation undertaken, generating significant theorems and/or implementation and/or analytical or empirical performance studies, realizing a paper appropriate for a journal such as TODS. In fact, the key role of archival journals is to publish solid, enduring, reproducible research in a format less constraining than workshop or conference proceedings (and benefitting from a more thorough, multi-pass review).

The following policy has been in place for many years, somewhat stemming the growth of the longest papers.

6. TODS will discourage excessively long papers (longer than 50 double-spaced pages including figures, references, etc.), and unnecessary digressions even in shorter papers. This is to motivate the authors to bring out the essence of their papers more clearly, to make it easier for the reviewers and readers, and to allow TODS to publish more papers in any given issue.

This policy states clearly the advantages of keeping journal papers focused, while retaining their positive qualities of completeness and rigor.

The fourth kind of paper is a survey of the literature or of practice. My previous editorial announced that TODS is now accepting certain kinds of directed surveys.

The final kind of paper has traditionally appeared in journals such as Information Processing Letters or ACM Letters on Programming Languages and Systems. These papers are also rigorous and contribute e.g. a faster algorithm or a theoretical insight that has not appeared elsewhere, yet have a more restricted scope than a conference or conventional journal paper. They are usually quite short, 5-20 pages, depending on the background they need to present.

The Editorial Board recently augmented the above maximum bound with the following policy which focuses on the minimum bound.

7. In a similar vein, TODS encourages shorter submissions, including even very short (say, five page) submissions. The primary criterion for acceptance is improving on the state-of-the-art in some significant way.

We are serious about encouraging shorter submissions. In fact, we are delighted to have received recently a submission of five pages and another of seven pages. We will inform referees of this new policy and will strive to reward succinct submissions with faster turnaround and favorable editorial decisions. We hope that authors who had decided their submission wasn't long enough for TODS will rethink that decision in light of this new policy.

On a related note, you can now submit your manuscript electronically, in a variety of formats. Just point your browser at the above-mentioned URL.

Richard T. Snodgrass, Editor-in-Chief

Author's address: Richard T. Snodgrass, Department of Computer Science, University of Arizona, P.O. Box 210077, Tucson, AZ 85721-0077, rts@cs.arizona.edu

This year marks the start of the second quarter-century of ACM Transactions on Database Systems. That in itself is not remarkable. In the last fifteen years, the number of serial titles across all disciplines has increased by 50%, to 164,400 today. New journals appear regularly, yet it is difficult to recall even a single journal that has been terminated. What is notable is the extent to which TODS has evolved since its founding to become the premier database journal.

Summarizing a citation analysis of database literature, considering over 100,000 citations, the web page http://www.acm.org/sigmod/dblp/db/about/top.html lists the top-cited papers and books. Thirty TODS papers appear on this list; 31 papers were from all other journals combined. TODS also dominated all conferences. The top-cited database paper of all time, Peter P.-S. Chen's "The Entity-Relationship Model," appeared in the inaugural issue of TODS.

TODS fares similarly well in a summary of estimated impact from the Research Index database (http://citeseer.nj.nec.com/impact.html), which ranked journals according to their average citation rate. TODS was judged the database journal with the most impact, appearing in the top four percent of the 800-odd journals and conferences analyzed.

These measures of impact correlate with journal accessibility (the causality probably goes both ways). It may surprise you that TODS has thousands more print and electronic subscribers than any other ACM Transactions. Of course, the journal (including all past issues) appears in the ACM Digital Library and is thus available to the many individual and institutional DL subscribers. TODS is also included in the SIGMOD Anthology and SIGMOD Digital Symposium Collection CDROM publications. 5000 copies were sent all over the world last year; another 5000 copies on DVD will appear shortly. These disparate media (print, web, CDROM, DVDROM), widely distributed, ensure that TODS articles are easily available to database researchers.

Due to the reputation and pervasiveness of this journal, taking over its editorship is a daunting proposition. The publication was founded by David K. Hsiao in 1976; Robert W. Taylor took over the editorship five years later and Gio Wiederhold five years after that. Won Kim served from 1992 until now, maintaining the excellence of the journal while increasing coverage of systems-related issues and decreasing the time to publication.

On behalf of the entire database community, I wish to extend our collective appreciation to the departing members of the Editorial Board, Serge Abiteboul, Yoshifumi Masunaga, Jeffrey F. Naughton, Moshe Y. Vardi and of course Won Kim, for their service to TODS over the past decade. It is due to their efforts that TODS is where it is today. And I welcome the following new members who have joined over the past year: Peter Buneman, Surajit Chaudhuri, Michael J. Franklin, Christian S. Jensen, Alberto Mendelzon and Z. Meral Özsoyoğlu. I would like to thank them and the continuing members of the Editorial Board for agreeing to serve. Their contact information is given on the masthead on the inside front cover. Finally, I welcome Curtis E. Dyreson who joins as TODS Information Director. Curtis has revamped the TODS web site at http://www.acm.org/tods and has plans for improvements down the road.

My challenge as Editor-in-Chief is to maintain TODS' position as the foremost database journal, while also ensuring that it is relevant and timely in this Internet age. It is my firm belief that with the glut of information on the web, much of it of dubious quality, the authoritative source of refereed papers on databases that is TODS will be even more important.

I am working closely with the Editorial Board to respond to the challenges confronting scientific publishing. Our first innovation is to solicit focused surveys on topics relevant to TODS. These should be deep and will sometimes be quite narrow, but should make a contribution to our understanding of an important area or subarea of databases. More general surveys that are intended for a broad-based Computer Science audience or surveys that may influence other areas of computing research should continue to go to ACM Computing Surveys. Brief surveys on recent developments in database research are more appropriate for ACM SIGMOD Record. TODS surveys should be educational to the database audience by presenting a relatively well-established body of database research.

These surveys are intended to summarize the "state-of-the-art", which the Random House dictionary defines as the latest and most sophisticated or advanced stage of a technology, art, or science: "a camera considered the state of the art in design". A similarly broad definition is employed here: surveys can summarize prior literature on a theoretical or systems research topic, or can explain approaches implemented in commercial systems. A survey of the former type summarizes the literature on a particular subject, presenting a new way of understanding how the papers in this literature fit together. A survey of the latter type summarizes the best industrial art and can be acceptable even if it represents no new contribution over what has been used in industry for years, if the paper's content is not to be found in the published literature. (Sometimes such practice is ahead of the published research; especially in such cases it is important for researchers to know about these engineering advances.) I thank José Blakeley for his persuasive advocacy of such surveys.

Due to a complex combination of factors, TODS issues have been coming out several months after their cover date. This tardiness concerns me greatly and must be addressed. I have been working hard on this, and have made some progress. In particular, TODS is in the first group of journals to adopt a new production process. In the past, accepted articles were converted from their submission format, LaTeX in most cases, to SGML for printing, which took considerable time and introduced errors. I am pleased to announce that from this point forward accepted articles will be typeset directly from their LaTeX source into PDF, for both the printed version and the version in the ACM Digital Library, provided that the source is prepared using the TODS style file (see the author information on the TODS web page for more details). This approach will result in fewer errors, faster production and thus a more timely publication. Moreover, since authors will know what the final version will look like, fonts and all, they can better fine-tune the appearance of their papers.

In closing, I ask that you keep three opportunities in mind. With its high quality, wide accessibility, short reviewing time, improved production process and small backlog (four to six months from notification of acceptance to appearance), there has never been a better time to submit to TODS. Secondly, I encourage you to consider submitting a survey article, whether on a theoretical topic, a systems topic, or previously undocumented approaches in implemented systems. Finally, TODS is above all a shared resource for the entire database community. I welcome your suggestions for how we can evolve this publication to better meet your needs.

Richard T. Snodgrass, Editor-in-Chief

Author's address: Richard T. Snodgrass, Department of Computer Science, University of Arizona, P.O. Box 210077, Tucson, AZ 85721-0077, rts@cs.arizona.edu

I wish to announce a number of changes to the Editorial Board of TODS and a new drive for increased relevance for the journal.

Editorial Board. Hector Garcia-Molina, H. V. Jagadish, and Paris Kanellakis, each having done a fine job as an associate editor, will conclude their three-years terms and leave the Editorial Board, in July, October, and September, respectively. It is a real pleasure to have worked with them; they have made tangible contributions to our field in their capacity as associate editors of TODS.

I have invited Rakesh Agrawal, Michael Carey, Michael Kifer, Richard Snodgrass, Patrick Valduriez, and Gerhard Weikum, whose three-year terms all expire during the second half of this year, to continue for another three-year term. They have all graciously agreed. Marianne Winslett, who started in 1994, will also continue her term.

To round out the Editorial Board, I have invited four new associate editors to join, starting in the second half of this year. The new associate editors are Katsumi Tanaka (Kobe University), Jeff Naughton (University of Wisconsin), and Moshe Vardi (Rice University). I look forward to working with them.

My term as editor-in-chief comes to an end in July 1995, but I have decided to continue for a second (and final) three-year term. The only reason for this decision is that the editorial policy that the new TODS Editorial Board and I established at the beginning has really been in effect for only the final year of my term. I want to see the new editorial policy take root because I firmly believe that the policy correctly addresses the changing nature of our field.

Every associate editor who joined the Editorial Board during my term has been dependable and competent. The Editorial Board has fulfilled the objective of completing each review cycle within about six months. The Board has also strictly limited the size of a submission to 50 double-spaced pages. Through these efforts, the backlog in the publication queue for TODS has virtually disappeared. So that papers may now appear in print within six months of acceptance. The Board has even accepted a few papers that have added substantial value to prior versions published in conference proceedings. The Board has also insisted on the integration of theory and practice, wherever possible, in all accepted papers.

Drive for Relevance. The emphasis on integration of theory and practice is an attempt to encourage authors of theory papers to consider applicability and/or implementability of the theoretical results, while encouraging authors of systems papers to reflect on the theoretical results that may have been used in building the systems and/or to offer suggestions on issues that may require theoretical treatment. The drive for the integration of theory and practice will receive even greater emphasis in the future.

There has been one very disappointing aspect. Most papers submitted to TODS over the years have been on data modeling, database architecture, database language, and the theories behind them. But TODS (and, for that matter, all archival scholarly journals that deal with data management) has failed to attract papers that report on database tools and new database requirements.

In particular, over the years, software vendors have developed and popularized a number of database tools such as graphical browsers of a database, fourth-generation language products, graphical user interface generators, and application generators, but the database research community has not been much of a factor in their development and refinement.

Further, while the database research community has been busy refining the traditional subjects of research, database vendors have quietly forged ahead in a number~of aspects of database architecture. These include stored procedures, replication service, workspace management in object-oriented database systems, integration of legacy database systems, and so on.

Today, the downsizing trend, and the resulting client/server computing and global networking, is forcing a reexamination of some major aspects of database technology, including support for multimedia data, distributed database architecture, and collaborative work. Further, although some progress has been reported, both from the research and the vendor community, more work is clearly needed in such areas as object-oriented database design methodology, databases for symmetric multiprocessors and parallel computers, scalability (i.e., keeping up performance as database size becomes massive), and authorization and multilevel security (especially for object-oriented database systems).

This is merely a sketch of the opportunities for the database research community. Ted Codd, with strong support from the System R and Ingres research groups, defined a primary focus for database research for the entire decade of the 1970s. During the decade of the 1980s, there was again one primary focus, but this time, it was to understand fundamental problems in relational technology and finding solutions for them. I, along with all the editors of TODS, believe that for the remainder of the decade and beyond, the primary focus of database research is to adapt database technology to the changing reality of the computing environment, thereby providing formal foundations for and correcting and solidifying the endeavors of the database industry in the areas above.

Although worthy research may be pursued (and has been pursued) outside the primary focus mentioned here, we feel that it is a useful beacon for the research community.

I, along with the Editorial Board, would like TODS to play an important role in realizing this vision. One concrete step is to publish invited papers on subjects that embody this vision, rather than wait for such papers to be submitted. Each member of the Editorial Board will invite and guide authors to write papers befitting the tradition of excellence that TODS has established over the years.

On behalf of the TODS Editorial Board, I thank the database research community for contributing to TODS by submitting papers and reviews over the years. As we strive to keep TODS relevant, we request your continued support.

Won Kim, Former Editor-in-Chief

The ACM Transactions on Database Systems (TODS) publishes original archival papers in the area of databases and closely related disciplines. The majority of the papers that have appeared in TODS address the logical and technical foundation of data management.

The international Editorial Board is composed of recognized experts in the various subareas of this field, all with a commitment to maintain TODS as the premier publication in this active field. Papers can be submitted directly to any of the editors. The Editorial Board maintains contact with ACM's Special Interest Group on Management and Organization of Data (SIGMOD), as well as with other societies, to encourage submittal of advanced and original papers. When appropriate, concise results may be submitted as technical notes; technical comments on earlier publications are welcome as well.

The existence of TODS has helped to define the field of database research. It encompasses the development, formalization, and validation of abstractions and models to describe database applications and the design and implementation methods for organizing and processing data on computer hardware.

TODS welcomes papers on a full range of subjects in the field of database research. They include, but are not limited to, data modeling, database language design, theoretical foundations of database languages, automatic query optimization and processing, access methods for storing and retrieving data, security and authorization, transaction management, concurrency control, backup and recovery, database performance tuning, and so forth.

TODS encourages papers that explore the above subjects in the context of large distributed networks of computers, parallel or multiprocessing computers, or new data devices (including data storage devices, data capture devices, and data presentation devices). TODS also encourages papers that describe emerging data-intensive applications that cannot be satisfied by the current database technology.

TODS welcomes papers that both lay theoretical foundations for database management and those that provide new insights into the design and implementation of large-scale database management systems, database application development tools, database access interface tools, and database connectivity tools for heterogeneous database systems. TODS also accepts papers that describe user and database administration experiences and issues in large-scale real-world database installations.

In terms of the ACM Computing Reviews (January 1995) classification, the primary area of TODS is all of area H (Information Systems), with a strong focus on subarea H2 (Database Management). Articles in subarea H4 (Information Systems Applications) would be appropriate only if there were a strong scientific basis in database technology.

Database systems may employ Specialized Languages (D.3.2) and unique Datatypes and Structures (D.3.3). Effective information retrieval and management require inferential power, and important advances are being made in this area. Topics Deduction and Theorem Proving (I.2.3), Knowledge Representation Formalisms and Methods (I.2.4), Learning (I.2.6), and Problem Solving, Control Methods, and Search (I.2.8) are all relevant when they are applied to large collections of data.

Since files provide a foundation for databases, the topic of Files (E.5) is covered as well, and many papers expand on operating system concerns, Storage Management (D.4.3), Reliability (D.4.5), Security and Protection (D.4.6), Organization and Design (D.4.7), and Performance (D.4.8). TODS publishes papers dealing with hardware systems for databases, but avoids those about specific devices, so that the topic Storage Hardware (B.3.2) is covered only in part.

TODS, however, does not publish surveys on purely descriptive material on database management systems or applications that employ database management systems. Evaluations of new concepts and their implementation, presented so that the results provide a foundation for further work, are of course encouraged, since feedback from practice is an essential aspect of the scientific paradigm.

Won Kim, Former Editor-in-Chief

Available only from the ACM Digital library in pdf.

Record-keeping and decision-making in industry and government are increasingly based on data stored in computer processable databases. Thus the need for improved computer technology for building, managing, and using these databases is clearly evident. This need is particularly acute in a complex society where the interrelationships among various aspects of the society must be identified and represented. The data which must be used to represent these relationships are growing more complex in nature and becoming greater in size. Furthermore, the increasing on-line use of computer systems and the proliferation and mass introduction of multilevel secondary storage suggests that future computer systems will be primarily oriented toward database management. The large size of future on-line databases will require the computer system to manage local as well as physical resources. The management of logical resources is concerned with the organization, access, update, storage, and sharing of the data and programs in the database. In addition, the sharing of data means that the database system must be capable of providing privacy protection and of controlling access to the users' data. The term data is interpreted broadly to include textual, numeric, and signal data as well as data found in structured records. The aim of ACM Transactions on Database Systems (TODS) is to serve as a focal point for an integrated dissemination of database research and development on storage and processor hardware, system software, applications, information science, information analysis, and file management. These areas are particularly relevant to the following ACM Special Interest Groups: Business Data Processing (SIGBDP), Information Retrieval (SIGIR), and Management of Data (SIG- MOD). TODS will also embrace parts of the Management/Database Systems and the Information Retrieval and Language Processing sections of Communications of the ACM. High quality papers on all aspects of computer database systems will be published in TODS. The scope of TODS emphasizes data structures; storage organization; data collection and dissemination; search and retrieval strategies; update strategies; access control techniques; data integrity; security and protection; design and implementation of database software; database related languages including data description languages, query languages, and procedural and nonprocedural data manipulation languages; language processing; analysis and classification of data; database utilities; data translation techniques; distributed database problems and techniques; database recovery and restart; database restructuring; adaptive data structures; concurrent access techniques; database computer hardware architecture; performance and evaluation; intelligent front ends; and related subjects such as privacy and economic issues.

TODS will not compete with proceedings of various SIG workshops. However, high quality papers from workshops, symposia, and conferences may be recommended to TODS upon prior arrangement with the Editor-in-Chief. The recommended papers will be refereed, and accepted papers cannot be published elsewhere. TODS is a quarterly publication with an initial annual size of about 400 pages. It is expected that articles will be longer than those appearing in workshop proceedings; the issues will probably average five or six articles.