Stevan Harnad
Department of Psychology
University of Southampton
Highfield, Southampton
SO17 1BJ UNITED KINGDOM
harnad@cogsci.soton.ac.uk
http://cogsci.soton.ac.uk/~harnad/
I should begin by defining some of the metaphors I use in this paper. By the "Gutenberg Galaxy" I mean the world of print on paper, and thus the "PostGutenberg Galaxy" is its successor, the virtual world of print on tape, disk and screen - and especially in the fibre-optic cables enmeshing the globe and transmitting bits at the speed of light everywhere. I also use the term "Skywriting," for the dissemination of the written word in the PostGutenberg Galaxy is very much like writing it all up in the sky, for everyone to see and to append their own scribblings onto, rather like the serial graffiti in public toilets, except on a galactic scale. Or perhaps a global Hyde Park, with the orations and cat-calls all delivered graphically rather than orally.
But perhaps the best simile for the Net is that it is like a clean nuclear weapon still being used mainly for children's games. For if you consider the only written corpus that matters to those who have taken the path of Learned Inquiry, then it is a fact that as of this day in late 1997, 99.9% of it is still available only as print on paper. There are a vast number of bits up there in the sky, but very few of them are the learned bits. And even those that are, are mostly hidden behind "firewalls" that keep out anyone who has not paid to view them.
Why not, you ask? People's writings were not given away gratis in the Gutenberg era either: why should skywriting be free now? After all, Learned Inquiry surely isn't literally (if one can be literal about a metaphor) a Hyde Park or Global Graffiti Board. But of course the authors of the learned serial literature - that means the contributors of everything that appears between the covers of the refereed scientific and scholarly journals to which research libraries must subscribe and on which all further research depends - are never paid a penny for their texts. Nor do they wish to be paid.
It's time for another metaphor, not a flattering one to scholars, but a handy way of understanding the anomaly of this state of affairs, unique in the Gutenberg Galaxy, where most writers are paid for their texts, in some cases very handsomely: When the scholar/scientist is wearing his learned-journal hat - for scholar/scientists may wear other hats too, when they are trying to write popular books or textbooks or magazine articles - when they are writing for their fellow-scholars in their specialised periodicals, the text they publish is much better thought of as an advertisement, rather than as anything analogous to normal fee- or royalty-based publication. It makes about as much sense for a learned author to restrict his work to those who have paid a fee to access it as it would for an advertiser to allow only those to read his adverts who have paid for the right to do so.
In his learned journal articles, a scholar is trying to make a contribution to knowledge. The only mark of having made a contribution to knowledge is to have one's work read, cited and built upon by one's fellow scholars. Otherwise one may as well not have done the work at all, pursuing instead a career with more tangible rewards. This is not to say that Learned Inquiry does not have material rewards. Few get Nobel Prizes or Field Medals, but promotion, tenure, travel -- plus the funds to perform further research (in the disciplines that rely on research grants) -- all of these depend on one's scholarly contribution; and the main measure of that contribution is one's learned serial publications. (I am agnostic about those fields of scholarship where it is books rather than journal articles that carry a scholar's intellectual goods, though I rather suspect that the story for esoteric learned monographs will be similar to the story for refereed journal articles.)
Although publish-or-perish bean-counting is still definitely with us, it was never a matter of publishing just anything, in just any journal: there is a hierarchy of learned journals in every field, and those at the top are rightly assigned a heavier weight in reviewing a scholar's contribution than the ones at the bottom. So if journal articles are like adverts, then their impact depends not only on the fact that they were published, but also on where they were published - in a street-corner flier or a Bergdorf- Goodman Catalogue. What the author actually wrote admittedly takes something of a back seat in all this, but it is assumed, usually correctly, that the real triage - the real "quality control," to use another revolting commercial descriptor that makes scholarship sound like garment-manufacture or the inspection of pork bellies, is performed by the referees and editors of the learned journals (and here the analogy with advertising breaks down). It is the peers of each realm of Learned Inquiry who decide what is worthy of publication, and where; refereeing is also called "peer review."
So, so far you have had a hint of why scholars publish at all: it's to make their work known to their peers, so others can build on it in the cumulative, collaborative enterprise that is Learned Inquiry. But it is also in order to get credit for their work in the cumulative, competitive enterprise called making a living. So far that's two reasons why journal articles are like advertisements rather than trade books or articles. Another reason is the "Impact Factor". This is a measure of how well-cited an author or journal is: it is the the number of citations divided by the number of articles. An Impact Factor of 1 means one citation per article. Even if that one citation is not a self- citation, it does not attest to much of an impact: it's rather like zero population growth, except it's zero knowledge growth.
It may surprise you to hear that an impact factor of 1 is actually quite high on average, because the average learned article gets no citations (and many may never even get read at all, except by the editor and referees). This does not imply that we should cut back the literature to only those papers that have high impact factors. First of all, you can't know a paper's impact factor in advance; second, some papers are late bloomers, their importance picked up only years after publication. But last, there is invariably a certain ratio of chaff to wheat in every field of human endeavour; rather than trying to cull it all out in advance, it makes more sense to have a hierarchy of journals in which peer review determines how high in the hierarchy a paper should go. The most rigorously reviewed journals are at the top, and it all grades down into a vanity press at the bottom. This is how it is today in paper, and this hierarchy must be duplicated on the Net.
But I'm getting ahead of myself again: we were talking about why it would even enter anyone's mind that Scholarly Skywriting should be given away, any more than terrestrial scribbling is. I've suggested an answer: what the scholar wants for his learned articles is eye-balls, not pennies, the eyes-balls of his peers, and the minds to which eye-balls are normally connected, from which future knowledge will flow. Let us call all of this "Impact" for short. It is because the learned author writes for Impact that he would much prefer that his bits reach every inquiring mind without any needless impediment.
In the Gutenberg era there was an unavoidable impediment: the technology of print on paper, though simpler and cheaper than the hand-copying of illuminated manuscripts (which was itself an improvement on the oral tradition), is nevertheless quite pricey compared to standing up on a soap-box in Hyde Park and proclaiming one's findings to one and all gratis. Nor is it very efficient, as the bits on paper must be awkwardly transported to all readers by some means, which is usually by carrying them to the subscribing library and then being met by an equal dose of legwork on the part of the would-be reader, who must fetch it from the library - if there is one, and if it subscribes to the journal in which the article appeared, or has an interlibrary loan service, or provides copyright-clearance-cleared photocopies, and if the author knows about the article, has the time to spare from research and reading to devote it instead to finding and fetching, and if he is lucky enough to be near a library that subscribes, or rich enough to subscribe himself.
These impediments are all real, and chief among them is the fact that the costliness of producing and distributing print on paper means that the publisher must make a sizeable investment in providing the technology and service if the author's work is to reach any eye-balls at all. So the admission ticket must continue to be collected at the door. As a consequence, it is a reality today that many scholars and potential scholars the world over have access to very little of the learned literature purely because of the economics and inefficiency of distributing and accessing learned serials as print on paper.
And let it not be thought that the only real cost of publishing paper journals is that of composition, type-setting, printing, and distribution (plus the cost of real advertising, and of collecting the toll itself). There is also the "quality control" I mentioned earlier: an editor must first evaluate submitted manuscripts by sending them to qualified specialists for peer review, then he must oversee the revisions and perhaps a second or third round of refereeing, before certifying it fit for publication and hence for scholarly consumption. Let us call this quality control for content. There is then also quality control for form, in the form of editing, copy-editing, mark-up and proof-reading. All of this costs money: not the all-important refereeing, paradoxically, because the peers do that for free, acting on a tacit "golden rule" - referee others as you would have them referee you. This is in part superstitious but in part also quite realistic. (If no one were willing to referee, none of the manuscripts submitted to refereed journals could appear -- or worse, all of them would appear, leaving us with no clue as to what was or was not fit for consumption.)
But even with the peers voluntarily stooping beneath the referee's yoke for free, the rest of the cost of quality control - that of implementing the peer review, copy-editing, etc - all calls for real money, over and above the paper-specific expenses. It is so as to recover all these costs (plus a fair profit) that the printed word needs to be sold rather than given away, not only in trade publishing but in scholarly serial publishing as well.
This is the point where I have to introduce the "Faustian Bargain," but let me say at the outset that this is not meant to be demonizing paper publishers at all; they, like the scholar, are joint victims of the technology and the economics of print on paper. For the Faustian Bargain is this: if you wish to immortalise your words at all, you will have to surrender your copyright in exchange, so your publisher can recover the substantial cost of getting your intellectual goods aboard the paper flotilla at all. The author must collaborate in denying access to his adverts to anyone who (or whose library) has not paid for them.
These are mixed metaphors - "Faust," "paper flotilla" - but we need the metaphor of the paper flotilla carrying the intellectual goods of Learned Inquiry in the Gutenberg Era so we can contrast it with launching them as Skywriting in the PostGutenberg Galaxy. (I especially like the french counterpart of Skywriting: "L'Ecriture Celeste.)
First, though, a little more about copyright, that forensic glue that binds together the shared interests of the trade author and the trade publisher in protecting their joint product from theft. In the trade sector of the Gutenberg Galaxy (which is to say, almost all of it) there was no conflict of interest, and this will not change in the trade sector of the PostGutenberg Galaxy either. Why should it? Whether you prefer to get my best-seller in paper or paperless form, I and my publisher still want to get paid for it.
But what about the scholar and his publisher? In the "Papyrocentric" Age, despite the conflict of interest inherent in restricting access to his work to those who have paid for it, the scholar/scientist had to collude in resolving the conflict in favour of his publisher (or rather in favour of the Mephistophilean Diktat of the Gutenberg technology) if he wanted to advertise and immortalise his work at all. This epoch is over; the distribution of the fruits of Learned Inquiry can at last be freed from the Faustian grip of the paper era. To see it directly, you must turn your eyes skyward toward XXX, the Los Alamos Physics Eprint Archive, created by Paul Ginsparg in 1991.
This remarkable virtual entity began as a decision by a small community of high energy physicists, about 100 researchers who had been sharing "preprints" of their work - manuscripts submitted, or soon to be submitted for peer review - in the old, expensive, inefficient way, by photocopying them and mailing them to one another. As they all had logins on the Net and were already in email contact with one another, it seemed sensible to exchange preprints electronically, without the trouble and expense of generating paper. It was only one step more to realise: why even send them electronically? Why not just archive them electronically and simply send notices of the contents of the publicly accessible archive to those on the email list or anyone else who was interested, and let everyone access whatever they want directly from the Archive?
Paul Ginsparg did just this, and within a few years, XXX had grown to encompass more than half of the current literature in most areas of physics, and it is still growing at an astonishing rate, with hundreds of papers archived weekly, tens of thousands of users the world over, and no doubt millions of "hits" of one sort or another, from the lookup of a reference to the download of a full "E-print," as they have come to be called. For XXX now contains not only the unrefereed preprint literature, but also the refereed, published reprint literature too.
At this point, Mephistopheles stirs and emits a muffled subterranean roar: "What? Published REprints? But what about the price I had exacted in exchange for immortality via paper? What about copyright!"
So what about copyright? Let's first remind ourselves of what copyright is intended to protect: the French expression for copyright is "droit d'auteur," "author's rights." What is it that copyright is protecting the author from? First and foremost, it is meant to protect him from the theft of the authorship of his text. If someone takes my text and publishes it as his own, and is awarded the Pulitzer Prize in my place, then he has indeed done me wrong, whether I am a trade author who seeks payment for his words or a scholar who does not. So there is nothing new about the Net insofar as copyright protection from theft of authorship is concerned.
It is true that it is easier to plagiarise a text when you can scoop up its bytes directly, and repackage them as your own, as you can in the electronic medium; but it is also easier to detect plagiarism in Skywriting than on paper. For one thing, there are automatic gremlins scampering around the Net day and night gathering and "caching" everything that appears, sifting its contents byte-for-byte in search of bit-strings that correspond to this interest or that. This is already true today, when most of what's up there in the sky is still junk. But it would be easy for an author or employer to send out such "knowbots" to search for bit-strings corresponding to samples from a particular text, to make sure it has not been plagiarised - or vice versa, as the case may be (authorship theft cuts both ways: think about it!).
Skywriting is also an excellent way of "date-stamping" text so as to establish primacy. There are still some technical date- and source- authentication problems that need to be ironed out, but suffice it to say that if I post my Fields-Medal-Winning proof on the Web today, there will be many more ways to confirm that it was indeed mine, and indeed posted by me today, than there would be if I had mailed paper copies to colleagues or to a journal - or if I presented it orally at a seminar, for that matter. To "publish" in the sky is to "go public" in a big way!
So let us not dwell on the authorship-protection function of copyright; that's a red herring: it will be at least as well protected on the Net as on paper. But there is another protection that copyright is meant to bestow, over and above protection from the theft of one's authorship of one's own text, and that is protection from the theft of the text itself. This is certainly what most people have in mind when they worry about the Net as a copyright risk.
Let me immediately say that a text anywhere on the Net is at great risk of being stolen, and that it will not be as easy to detect that kind of theft as it will be to detect the theft of authorship. This is where the "firewalls" I mentioned earlier come in - although that is a double metaphor: not only are network firewalls not really walls of fire, but the term is normally used to refer to barriers against public entry into a private network, rather than to barriers against unpaid access to digital texts. But as you see, these two come to much the same thing. There will be reliable firewalls that will block access to texts until they are duly paid for, perhaps with digital cash, but I hope it is clear to you by now that this kind of firewall protection is not at all what the authors of learned serial articles need or want. The authors of the preprints and reprints in XXX are archiving them there precisely so as to remove all firewalls obstructing the free flow of their bits to any mind that may wish to eyeball them, anywhere, at any time.
So copyright protection from theft of text is literally moot in the special literature we have been discussing: copyright cannot and need not be invoked to protect an author from something against which he wishes no protection: Le droit d'auteur n'est pas la pour me proteger contre moi-meme!
What about the publisher? The reason scholars entered into the Faustian Bargain in the paper era was that if they did not assign copyright to their publishers then the texts could be stolen, the publishers would fail to recover the costs of publishing them, and so they could not be published at all. If this is no longer true in the new medium, what has changed to make it so?
What is the difference in cost between a paper page and a skywritten page? It depends on whom you ask. Paper publishers will tell you that the difference is not that much, because the lion's share of the cost of publication is the price of producing that first page you print; the rest of the copies of that page cost little more whether you print ten of them or ten thousand; and that 1st-page cost, which you will have whether you go into paper or into bits, represents 75% of the current cost of paper publication; the saving from going electronic is a mere 25%, which does not sound like the basis for any revolutionary change in economic model. Moreover, because for the time being there still seems to be a demand for that much maligned paper edition, the cost of producing both a paper and an electronic incarnation of a journal actually increases the total cost to 125% per page.
On the basis of this reckoning, most paper publishers are proposing something along the following lines for the future: for now, as there is still a demand for the paper edition, it can be subscribed to at the usual price; a subscription for both the paper and the electronic edition will cost a little more, and the electronic edition only, a little less. This way publishers are ready for the transition whenever we are, and everything proceeds pretty much as it has been doing until now. The author's hopes of giving his texts away are alas quashed. As in the Gutenberg era, cost-recovery necessitates charging admission at the door. The Faustian Bargain continues to be the only deal in town.
But if instead of asking a paper publisher how much it costs to produce an electronic-only version of a learned journal you ask some of the brave souls who have launched electronic-only journals, you will hear that the savings are more like 75% per page, and probably even better than that. What is responsible for the difference in arithmetic? I think it is that the paper publishers are reckoning the page costs the wrong way. They are wrapping all their current expenses into the page price, and then subtracting the line items that are specifically for generating the hard copy. This means they are retaining the costs of an infrastructure that is purpose-built for producing and distributing print-on-paper, and merely subtracting print-specific expenses. This is like a horse-drawn transport service calculating that if they switched to horseless carriages all it would save would be the price of the hay!
Electronic-only journals that have started bottom-up, without any of the papyrocentric infrastucture, are reporting much lower page costs, but unfortunately they are also having trouble filling those pages, because the hard currency for recognition and advancement in Learned Inquiry is still paper, for the simple reason that all the established journals happen to be paper ones. So although it all sounds good in theory that free electronic journals are optimal and inevitable, the optimal and the inevitable still look a long way off.
I will now close with a breathless sketch of a viable transition scenario to the optimal and inevitable outcome in learned serial publishing, followed by a brief coda about the most revolutionary feature of Learned Skywriting, one that remains to be explored.
I've made a subversive proposal for hastening the day: What is needed is a way that scholars can have their cake and eat it too: They should continue to submit their papers to the prestigious paper journals of their choice, but they should at the same time archive them publicly on the Web, first as unrefereed preprints and then, once they have been reviewed, revised and accepted for publication by the paper journal, as refereed reprints. The archiving can be done in either or both of two ways: on their local Web Server or on a centralized Eprint server, mirrored world-wide, like XXX, preferably both, to help the search gremlins gather, organise, and index it in more uniform ways. These Eprint archives should be strongly supported by institutions and research funding agencies.
Unless there is something unaccountably different about physics, the other disciplines should follow the same pattern and come to use the Archive as the locus classicus for accessing the journal literature. As the Archive comes into its own, libraries should at last begin getting the green light from their users to start cancelling journal subscriptions. The sooner publishers see the handwriting on the wall - or in the sky, as the case may be - they should start to take steps for some radical perestroika if they wish to continue publishing learned serials.
The scenario branches here for those publishers who do elect to change and those who don't: Those publishers who do not restructure themselves, who persist in trying to use the subscription or site license or pay-per-view model (S/SL/PPV) for cost recovery, especially if they attempt to use submission policy and copyright as a way of preventing their authors from publicly archiving their preprints and reprints, respectively (as many are doing now), they will simply lose their Editorial Boards, who will emigrate to the Web on their own, under user pressure (after all, they are us!), and will reconstitute themselves as electronic-only journals, with or without the old brand name, so as to recover the much lower page costs through author-end page charges, instead of S/SL/PPV, all of which block access to the reader.
The page-charges will at first be paid entirely by government and university subsidy. This will amount to less than $50 per page, to pay for the only remaining costs of learned ejournal publication: the price of implementing peer review and providing editing, copy-editing and mark-up - in other words, the cost of quality control of form and content. As the culture changes, these page-charges will be covered as part of the cost of supporting scholarly-scientific research itself. The preferred scenario is for the present paper publishers to restructure themselves on the new model, ensuring continuity of the known and trusted brand-name journals. Either way, learned serial publication will become an overlay on the Learned Eprint Archive, authenticating the peer-reviewed sector with their imprimatur, but with the Archive itself as the point of access.
In exchange for all this, the learned serial literature will be available everywhere, for everyone, for free, forever, as it always would have been, but for the tyranny of print on paper in the Gutenberg Era.
Now a coda about peer review versus peer commentary:
I doubt I'm the first to notice that the World Wide Web today is like the Wild Wild West a couple of centuries ago: There's the same sense of limitless space where the shackles of the old order can at last be cast off. And of course there's no scarcity of cowboys. This demography and this atmosphere have persuaded many scholars and scientists that the Net is not a suitable medium for serious work, apart from exchanging one-on-one email, perhaps.
This Wild-West ethos has given rise not only to a touching but unrealistic animus against charging money for anything on the Web, but also, when it comes to publication, a rejection as censorship of any form of refereeing or editing. These well- intentioned souls have suggested that prepublication peer review is best replaced by post-publication peer commentary: let everything appear, and then let readers vote with their eyeballs and skywriting.
I have edited for 20 years a paper journal, Behavioral and Brain Sciences (BBS), published by Cambridge University Press. BBS specialises in publishing important and controversial "target articles" in the biobehavioral and cognitive sciences. Each article is co-published with up to 30 peer commentaries, plus the author's response. This feature has earned a whopping Impact Factor of 15 for BBS, higher than any but the widest spectrum interdisciplinary science journals like Science and Nature. BBS's Impact Factor attests to the value the scientific community finds in having work treated in this interactive way, and its "open peer commentary" feature has since been accorded the sincerest form of flattery as it has been imitated by several other journals, just as BBS itself was modeled on Current Anthropology, the first journal of open peer commentary.
Peer commentary, however, is no substitute for peer review, and indeed, BBS is a very rigorously peer-reviewed journal, with a high rejection rate and each accepted article going through several rounds of revision and re-refereeing before it is allowed to run the gauntlet of open peer commentary.
Every editor of a learned journal, commentary journal or not, is in a position to sample what the literature would have looked like if everything had appeared without review. Not only would a vanity press of raw manuscripts be unnavigable, but the brave souls who had nothing better to do than to sift through all that chaff and post their commentaries to guide us would be the last ones to trust for calibrating one's finite reading time.
Peer commentary is a superb supplement to peer review, but it is certainly no substitute for it. But if you now imagine the refereed literature appearing in the skies freely available for all, then of course commentaries - both refereed ones and informal unrefereed ones - could enrich that literature at a speed and on a scale that the collaborative, cumulative corpus of Learned Inquiry has never before known. The wonders of skywritten peer commentary, however, are a topic for another lecture.
BIBLIOGRAPHY
Ginsparg, P. (1994) First Steps Towards Electronic Research Communication. Computers in Physics. (August, American Institute of Physics). 8(4): 390-396. http://xxx.lanl.gov/blurb/
Harnad, S. (1990) Scholarly Skywriting and the Prepublication Continuum of Scientific Inquiry. Psychological Science 1: 342 - 343 (reprinted in Current Contents 45: 9-13, November 11 1991). ftp://ftp.princeton.edu/pub/harnad/Harnad/HTML/harnad90.skywriting.html
Harnad, S. (1991) Post-Gutenberg Galaxy: The Fourth Revolution in the Means of Production of Knowledge. Public-Access Computer Systems Review 2 (1): 39 - 53 (also reprinted in PACS Annual Review Volume 2 1992; and in R. D. Mason (ed.) Computer Conferencing: The Last Word. Beach Holme Publishers, 1992; and in: M. Strangelove & D. Kovacs: Directory of Electronic Journals, Newsletters, and Academic Discussion Lists (A. Okerson, ed), 2nd edition. Washington, DC, Association of Research Libraries, Office of Scientific & Academic Publishing, 1992); and in Hungarian translation in REPLIKA 1994; and in Japanese in "Research and Development of Scholarly Information Dissemination Systems 1994-1995. http://cogsci.soton.ac.uk/harnad/Papers/Harnad/harnad91.postgutenberg.html
Harnad, S. (1992) Interactive Publication: Extending the American Physical Society's Discipline-Specific Model for Electronic Publishing. Serials Review, Special Issue on Economics Models for Electronic Publishing, pp. 58 - 61. http://cogsci.soton.ac.uk/harnad/Papers/Harnad/harnad92.interactivpub.html
Harnad, S. (1995b) Universal FTP Archives for Esoteric Science and Scholarship: A Subversive Proposal. In: Ann Okerson & James O'Donnell (Eds.) Scholarly Journals at the Crossroads; A Subversive Proposal for Electronic Publishing. Washington, DC., Association of Research Libraries, June 1995. http://cogsci.soton.ac.uk/~harnad/subvert.html
Harnad, S. (1995c) Interactive Cognition: Exploring the Potential of Electronic Quote/Commenting. In: B. Gorayska & J.L. Mey (Eds.) Cognitive Technology: In Search of a Humane Interface. Elsevier. Pp. 397-414. http://cogsci.soton.ac.uk/harnad/Papers/Harnad/harnad95.interactive.cognitio n.html
Harnad, S. (1996) Implementing Peer Review on the Net: Scientific Quality Control in Scholarly Electronic Journals. In: Peek, R. & Newby, G. (Eds.) Electronic Publishing Confronts Academia: The Agenda for the Year 2000. Cambridge MA: MIT Press. Pp. 103-118. http://cogsci.soton.ac.uk/harnad/Papers/Harnad/harnad96.peer.review.html
Harnad, S. (1997a) How to Fast-Forward Serials to the Inevitable and the Optimal for Scholars and Scientists. Serials Librarian 30: 73-81. (Reprinted in C. Christiansen & C. Leatham, Eds. Pioneering New Serials Frontiers: From Petroglyphs to CyberSerials. NY: Haworth Press.) http://cogsci.soton.ac.uk/harnad/Papers/Harnad/harnad97.learned.serials
Harnad, S. & Hemus, M. (1998) All Or None: There Are No Stable Hybrid or Half-Way Solutions for Launching the Learned Periodical Literature into the PostGutenberg Galaxy. In: I. Butterworth (Ed.) The Impact of Electronic Publishing on the Academic Community. London: Portland Press (in press) http://cogsci.soton.ac.uk/harnad/Papers/Harnad/harnad97.hybrid.pub
Odlyzko, A.M. (1995) Tragic loss or good riddance? The impending demise of traditional scholarly journals, International Journal of Human-Computer Studies (formerly International Journal of Man-Machine Studies), 42 (1995), 71-122. Condensed version in Notices of the Amercan Mathematical Society, 42 (Jan. 1995), 49-53. Available at URL http://www.research.att.com/~amo/doc/recent.html