The panelists have been invited to put forward their visions of the future of mathematical communication and publishing. Before trying to do that, I would like to tell you about a recent initiative that I and many others believe will be a major determining factor in the way the future plays out in these matters. Let me begin by reminding you of an important development in our sister science, physics, that had its origins early in this decade.
The LANL Physics E-print archive was started by Paul Ginsparg in 1991. It currently has more than 75 thousand e-prints, accruing at a rate of over 25 thousand per year. It is maintained by a full-time staff and offers many technical conveniences, including an easy to use form-based World-Wide-Web interface, an automatic TeX compilation system, e-mail notification, excellent search facilities, and a network of mirror sites in 16 countries. It has many tens of thousands of users every day, and constitutes a major change in the way physicists communicate their research results. Many experts who have looked carefully at the archive believe it is a significant step forward, and see it is as a first step towards even more far-reaching changes in scholarly communication across all fields of knowledge.
The maintenance costs for the archive are covered by grants from the US Department of Energy and the NSF. Surprisingly, these costs are only the same order of magnitude as those for the maintenance of a single good mathematical research library, and moreover adding the mathematics and computer science archives (see below) does not add significantly to the cost. It is these low costs and good scaling behavior that allows the archive to serves the whole world without access fees.
For several years, David Morrison and Joe Christy have worked with Paul Ginsparg and the LANL staff to provide a similar e-print archive of mathematical papers, but only in four specially chosen mathematical subject areas. At the beginning of 1998 it was decided to expand this to a complete mathematical e-print archive. This was accomplished by moving to LANL several other e-print archives on the Internet (e.g., Several Complex Variables, MAGNUS, Logic E-prints, Commutative Algebra, the Banach Archive, and LACES), combining them with the four mathematics subject archives already there, and adding other subject areas to cover the full range of mathematics. Although relatively new, this LANL Mathematics E-print archive already has 5,500 e-prints, and it is increasing by over 200 each month. A Steering Committee, consisting of a group of about fifteen mathematicians, chaired by Dave Morrison, was formed to direct this ongoing expansion in coordination with the LANL archive staff at Los Alamos. (Dave was originally going to be on this panel, but he was unable to make it. I am one of the members of the Steering Committee, and I agreed to fill in for Dave. I would like to make it clear that when I say ``I'' or ``we'' below I am speaking for the Steering Committee.)
In August of 1998 a parallel effort was initiated in computer science with a steering committee formed in cooperation with an ACM Publications Committee.
We expect that over the next several years LANL will become the home of a substantial fraction of the entire primary mathematical literature, and I would like to invite each of you to get started using it by submitting at least one of your research articles in mathematics to LANL, if only to learn the system. Articles that have already been published are welcome, and if you feel that an older article of yours is of unusual significance, please do make it available at LANL. (If you have signed away your copyright, you may need permission to do this, but enlightened publishers---including the AMS---no longer require you to ask their permission.)
Instructions for uploading articles are available at: http://xxx.lanl.gov/help/submit
Another member of the Steering Committee, Greg Kuperberg, has created a very user-friendly ``front-end'' to the LANL mathematical e-print archive at: http://front.math.ucdavis.edu/ , and you may find ``The Front'' a more convenient interface to LANL.
Anything you contribute will give your work immediate and significant visibility, since thousands of mathematicians regularly browse the archive. More importantly, your presence will encourage your closest colleagues to contribute also; it will help establish the use of LANL e-prints in your research areas. The more we all work for the success of the new system, the better.
For these same reasons, I would also like, on behalf of the Steering Committee, to issue an invitation to those responsible for the distribution of the ICM Proceedings volumes to make them available at LANL.
Some people worry that by going electronic and trusting the priceless heritage of our mathematical literature to a centralized database, we open ourselves to the danger of its destruction in some modern day catastrophe that would be a high-tech reprise of the burning of the great library at Alexandria in ancient times.
While no system can guarantee perfect reliability, there are good reasons to believe that LANL, with its careful security procedures that include frequent backups and highly redundant data storage, approachs the safety of the current paper-based system. The seven year on-line record of LANL has been exceptional, with essentially no unexpected downtime (even during the continual hardware and software upgrades), nor has there ever been a loss of data or a broken or changed URL. The 15 country mirror network, updated daily, is equally well-maintained, and helps to guarantee continuous world-wide access to the archival database even through intermittent or lost bandwidth.
Another often mentioned danger is that some change in storage format or media will leave us with our database of the mathematical literature intact but unreadable. But it is only rarely used or poorly maintained databases that are at risk when electronic formats change. Large, widely used databases, such as the LANL archive, can be and have been translated to new formats with mininal cost and effort.
This leads me to three crucial decisions made in the design of the LANL Mathematics E-print Archive. They concern universality, permanence, and centralization. First, the LANL archive covers all the subfields of mathematics. Secondly, the LANL archive is not a preprint archive, but rather an e-print archive. That is, articles posted to LANL are expected in almost all cases to remain there in perpetuity. And finally, all the articles in the LANL archive reside physically at LANL (as well as on the various mirrors).
Other electronic archives of mathematical papers generally differ in one or more respects. Often they are devoted to a special subject area within mathematics. Also, they are frequently preprint archives, and it is expected that articles will be removed after publication in some paper journal. Finally they are often decentralized; that is, the papers exist physically on many hundreds of different individual or departmental sites, and the central archive is really only a database of ``hyperlinks'' or pointers to these papers.
Essentially all e-archives that have based their design on this decentralized approach find it to be a serious Achille's heel that eventually renders them dysfunctional. One obvious problem is that the various sites that make up the network are mutually incompatible in their organization and formatting of data. But there is an even more serious problem. A database of pointers becomes useless for most purposes if a large fraction of the pointers are dangling. And this is just what tends to happen. Departmental and individual servers often are poorly maintained for financial reasons, and the archives they host often go offline either temporarily or permanently when a disk crashes or the graduate assistant in charge leaves without having documented the system he or she has built. And since preprints are considered transitory and expendable, no one worries much about updating them when media or formats change.
A guiding principle of the LANL Mathematics E-print Archive is that it is completely open, and freely available for all to use in ways that add scholarly value. Let me illustrate this by an example that demonstrates how using the archive can improve the economics of mathematical journal publishing.
Using a model referred to as an ``overlay'' journal, publishers will be able to produce electronic journals at minimum cost and so with minimum fees. Papers will be received, refereed, and accepted or rejected in the time-honored manner, but completely electronically. An author wishing to submit a paper to an overlay journal will initially post it to LANL math e-prints, and send a message to an editor with a link to the location of the paper on the archive. The editor in turn will forward that link to the referees. Accepted papers would constitute the journal; they would be accessible from a special journal site on the web, but would continue to be accessible directly from the LANL archive.
Other electronic journals will choose to go beyond this bare-bones approach, adding value in various ways (for example by copy-editing authors' manuscripts). This will incur added expense, but by avoiding the costs of printing, binding, and mailing, these electronic journals too will be considerably less expensive to produce than traditional paper journals.
Even overlay electronic journals will continue to have some costs associated to them, and so require financial support. But ``going electronic'' will contribute substantially to the most essential goal of publication, namely to keep the entire mathematical primary literature freely available to scholars everywhere. And insofar as we mathematicians leave our papers on the LANL e-print archive, no one will be prevented from accessing them.
Fundamental changes in scholarly communication in mathematics are on the horizon, and we believe that mathematicians need to intervene now to ensure that the system that emerges meets our needs. Historically, the mathematical literature has been maintained by an array of publishers, and universal indexing (provided by Math Reviews and Zentralblatt) has only come afterwards. With the Internet, it is possible, and in our view important and desirable, for the mathematical community to establish a free, universal, primary database of e-prints, that will allow rapid access to the literature from anywhere in the world.
Let me close with some quotations from Paul Ginsparg:
No one knows precisely what form our research infrastructure will take two decades from now, much less a century from now, but we do know that the role of many of the traditional participants must change, and this provides exciting possibilities for shaping this future. In particular, we do not expect that researchers themselves will have to play the dominant role in maintaining the global knowledge network of the future, now in its first stages of incipient development.
If we the researchers are not writing with the expectation of making money directly from our efforts, then there is no earthly reason why anyone else should make money in the process (except for a fair return on any non-trivial ``value-added'' they may provide) ...
One possibility is that some consortium of professional societies and institutional libraries will ultimately acquire the technical competence to provide umbrella sponsorship of the global raw research archive. That is my most optimistic outlook for the future, with the new technology ironically allowing those traditional players from a century ago to return to their dominant role in support of the research enterprise.