This is an expanded version of:
The final state toward which the learned journal literature is evolving in the age of networked hypermedia is as inevitable as it is optimal: Sooner or later, the entire corpus will be fully and freely accessible and navigable from the desk of any thinker in the world. The effects of this on the scope and pace of Learned Inquiry itself will be revolutionary, comparable only to the impact of three prior cognitive revolutions: the advent of speech itself, then writing, then print. Learned Inquiry, always communal and cumulative, will not only be immeasurably better informed, new findings percolating through minds and media almost instantaneously, but it will also become incomparably more interactive, with collaborative, creative, critical, and self-corrective cycles accelerable, potentially, almost to the speed of thought (Harnad 1991).
So much for the inevitability and optimality of it all: When will it come to pass? It could already have happened yesterday; everything is in place, technologically speaking, and further resources are poised to follow suit whenever we are ready. But there is no second-guessing human nature; old habits find no end of rationales for perpetuating both themselves and the status quo, even when, like a superannuated alpha wolf, they are far past the stage of being able to defend their status by anything but a menacing stare!
We will keep hearing solemn worries about the (1) preservability of texts in the new medium. In reality, however, nothing is simpler and more natural -- particularly as more and more of our intellectual goods take flight from the paper flotilla into the digital skies, and our vested interests up there become ever more collective -- than to arrange jointly for their continuous, systematic uploading and upgrading pari passu with ongoing developments in the medium: Bits are bits, and far more readily transferable than flecks of ink. Moreover, local preservation efforts scale up to an ever-growing set of eggs, particularly if they are all kept in the same basket (as I will recommend below).
The (2) less-than-optimality of even the most advanced current screen-reading will no doubt continue to be invoked as grounds for holding the new medium at arm's length. What this tireless plaint fails to reflect on is what proportion of our intercourse with the periodical literature really calls for the _kind_ of reading (bath-, bed-, beach-based) that is usually envisioned in this context, as opposed to the desk-based searching, skimming, spot-checking, citation-tracing, and active cut-pasting and quote/commenting that is the mainstay of Learned Inquiry. Besides, there's still the desk-side printer for whenever a huggable copy is needed; and long after the printer is obsolete, there will still be the option of micro-thin "virtual papers" that can simulate any papyrocentric feature (optimal or counter-optimal) after which habit hankers, from portability, pliability, and thumbability to even the musty smell of shelf-worn vellum.
But the biggest brake on progress is still surely the reluctance of authors to entrust their work to a new, unproven medium in place of the one that has served them faithfully for centuries. Nor is their resistance based primarily on worries about preservability or readability: Authors are concerned about (3) quality control (will the new medium be as reliable and rigorous as paper?), (4) credit (will it bring recognition and advancement as paper did?), and (5) plagiarism (will it make work more vulnerable to theft?).
The answers here are just as transparent: (3) Peer review, scholarship's classical mechanism of quality control, is medium-independent: it can be implemented in the air as readily as on land or sea (Harnad 1996). (4) Credit, likewise medium-independent, follows the best work and workers, it does not lead them (though authors and media can be like horses and water); kudos is ready to be assigned to skywriting as soon as authors are ready to redirect their pens heavenward. (5) Copyright too is medium-independent; and whatever increased power the Net may provide for stealing texts, it more than matches with its power for tracking them down (Harnad 1997c, Bachrach et al. 1998).
Yet, despite these 5 prima facie obstacles, things may look to be moving along quickly enough: The latest (7th) Edition of the Association of Research Libraries (ARL) Directory of Electronic Journals, Newsletters and Academic Discussion Lists ( http://www.arl.org:591) listed over 3,400 electronic serials in 1997, twice as many as in 1996, 29% in science, 28% in social science, and 14% in arts and humanities, 1,002 of them refereed (compared to 47 in 1996). This 2000% increase comes largely from the fact that more and more existing paper journals are now making electronic versions available for a fee: ARL listed 708 (compared to 168 in 1996), but this was already an underestimate, as last year's _Nature_ review of new electronic journals (389, 137-138; 1997) had already listed over 2000 such electronic doppelgaengers.
New, e-only journals, mostly free, are also being founded at a rapid rate. (Alas, many of these are also foundering almost as quickly as still newer ones appear. In this one respect the economic restraints on paper publication were probably beneficial: More intellectual and ergonomic effort and weight had to be put behind new paper titles to justify the high start-up costs, thereby fortifying if not guaranteeing their chances of survival. The absence of this economically dictated quality filter on new e-journal start-ups is probably a retardant, overall, on the speed of the transition to the PostGutenberg Galaxy, because it reinforces people's fears that there is something unreliable about the medium.)
But a twentyfold jump in the number of refereed journals on the Net from 96-97, even if based more on 2nd incarnations of existing paper journals than on new launches, is still progress toward the optimal and the inevitable, is it not? After all, even a mere tenfold leap from 98-99 would put the number well over the 6500 refereed journals indexed by the Institute for Scientific Information and near the 14,000 indexed by Ulrich's International Periodicals Directory. Yale's Ann Okerson (personal communication), Founder of NEWJOUR (http://gort.ucsd.edu/newjour) the e-list of new e-journals, estimates that there are now over 8000 refereed on-line journals. That is beginning to sound as if the core corpus may soon be in the sky for one and all.
For one and all? Recall that the optimal and inevitable was described as "fully and freely accessible," whereas we are observing a proliferation of fee-based incarnations of paper journals on the Net, with financial firewalls separating them from one another as well as from the user. This fee/free issue -- and it is a crucial and controversial one -- needs to be faced head-on.
The trick is to present it even-handedly: The advocates of the status quo hold that things are progressing just as they ought. Both paper and on-line versions of journals are obviously still in demand, so let supply and demand orchestrate the transition. Keep providing both versions -- the usual offer is the on-line version only for somewhat less than the print version only, and both for somewhat more, but it averages out to the current price structure, regardless of where and when the demand balance shifts -- and then phase out paper if and when the market demands. There has been some consideration of pricing mechanisms other than subscription (site-license and pay-per-view being the two other horses in the "trade troika": S/SL/PPV), but the overall revenue per published page of journal has been held more or less fixed in the reckoning.
What about the cost? Here estimates have diverged: Paper journal publishers, diversifying toward hybrid paper/on-line editions, estimate that costs per page increase (particularly as value is added to the on-line pages in the form of hyperlinks, multimedia, etc.), so there is certainly no basis for price reduction. What about when demand shifts toward on-line-only? It is a truism in publishing that expenses must be reckoned in terms of _first-page_ costs: It is all the work (writing, refereeing, editing, composition, mark-up) that goes into the first page that represents the lion's share of the expense. Printing and distributing multiple copies of that 1st page represents at the very most 30% of the page cost, and if and when that 30% saving can be made, publishers will be happy to pass it on to subscribers; for now, though, costs are up, not down (e.g., Garson 1997).
The other side of the story comes from the minority of journals that have started as on-line only. There are as yet no equivalent pairs of paper-only and on-line-only journals, exactly comparable in terms of submission rate, acceptance rate, pages per year, subject matter, readership, authorship, impact factor (citation ratio), and the all-important "prestige" factor, but such comparisons as have been made (e.g. Odlyzko 1998) suggest that on-line journals are managing for at least 70% less, and that these economies could even be improved upon as global Archives come into the picture, scaling down the first page costs to those of implementing (not performing) peer review (just as authors write gratis, referees referee gratis, and always have), plus editing/copy-editing and mark-up (much of that becoming increasingly off-loadable to authors too, as friendly HTML and eventually SGML authoring tools appear and pressure for their disciplined use by all authors grows).
A great deal rides on this discrepancy in estimates for the true cost difference between paper and on-line-only pages: If the saving is really only 30% or less, then S/SL/PPV is indeed the only way to recover costs, and free access, though still optimal, becomes unattainable instead of inevitable -- absent a massive subsidy (from whom? to whom?).
But if the savings are the 70% or more that the new e-only journal publishers are experiencing, then the optimal is not only inevitable, but within reach: Currently, that 70% is paid in large part by Library subscriptions. Simple arithmetic shows that if the page charges for the remaining 30% were paid in advance, at the author's end, out of (and for the sake of) those savings, not only would Institutional Libraries be much better off, but the world Learned Community -- authors and readers alike -- would be fast-forwarded straight into the optimal (Harnad 1997a).
Unfortunately, the world Learned Community (as our earlier "from whom?/to whom?" question suggests) is not a collective with any coherence or clout. Moreover, even if Universities rechannelled all savings from Library Serials cancellations to Faculty publication budgets, there would not be a balanced quid pro quo, because highly research-active Universities (the net page-providers, currently) would then face the biggest costs, whereas less research-active Universities (the net page-consumers, currently) would get a free ride. Not that there would be anything wrong with that, as long as there were enough funds to keep it all aloft. (And there might well be, because (i) the research-active Universities also tend to have the biggest serials collections, (ii) the 70% savings still leave considerable room for manoeuvring, and (iii), research productivity being the lifeblood of Learned Inquiry, and publication being its most measurable life-sign, other sources of research support, university as well as governmental, would find it natural to wrap into the funds for _conducting_ the all-important research the relatively minor marginal cost of _reporting_ it.)
But as long as restructuring depends upon a collective -- and an interdisciplinary, interinstitutional, and international one at that -- the optimal and inevitable may have a long wait. What is needed is some incentive to take matters into one's own hands as individual members of the Learned Community. There is a way, and it would allow individual scholars to have their cake and eat it too. The proposal is simple, and subversive (Harnad 1995b in Okerson & O'Donnell 1995): All authors should continue to entrust their work to the paper journals of their choice. But if, in addition, they were to publicly archive their pre-refereeing preprints and then their post-refereeing reprints on-line on their Home Servers, for free for all, then the de facto practises of the reader community would take care of the rest (irrespective of their reservations about bed/bath/beach reading); library serial cancellations, the collapse of the paper cardhouse, publisher perestroika, and a free for all, e-only serial corpus financed by author-end page charges would soon follow suit.
A centralised variant of this subversion scenario, http://xxx.lanl.gov, has already passed the point of no return in Physics and some allied disciplines in the form of Paul Ginsparg's (1994, 1996) U.S. NSF- (National Science Foundation) and DOE- (Department of Energy) supported Physics Eprint Archive at Los Alamos National Laboratory; as history will confirm, he single-handedly set the world Learned Community on its inexorable course toward the optimal and the inevitable in August 1991. Hence 1998 e-journals are not the right beans to count for e-future-casting: The number of new papers being deposited by authors at xxx (that subversive sapphire masked by all the look-alike cyber-smut sites) is now 100 per weekday, with the number of "hits" by readers now 65,000 daily and both figures still growing linearly http://xxx.lanl.gov/cgi-bin/todays_stats. The real measures of the radical new way the world physics community is accessing its research literature are these, not the rate at which the fee-based on-line doppelgaengers are being put on the market.
In response to this reality, the forward-looking American Physical Society, publisher of some of the most prestigious physics journals in the world, has agreed to collaborate with xxx, which is already the de facto locus classicus for much of the physics literature: Manuscripts will be submitted to APS journals through xxx for refereeing; the final, refereed, edited version can then also appear through xxx, but with APS certification. These are the advantages of eggs-in-one-basket centralisation. Backed by APS and NSF/DOE (and backed up at numerous mirror sites the world over), the "preservation" of this collective corpus is assured.
But the nature of cyberbaskets is such that there is always room for more eggs. (The entire refereed journal corpus across all disciplines is the size of the flea on the tail of the dog, compared to, say, the storage capacity and bandwidth planned to serve the demand for entertainment video. Even their 14,000 refereed serials represent only 1/10 of the magazines covered by Ulrich's.) Mathematics, computer science, and cognitive science are already joining the disciplines served by xxx; there is no reason whatever why xxx should not go on to subsume (or subserve, rather) all the rest of the learned serial literature too; economies of scale suggest that this may even be the most efficient and direct path to the optimal; and a transition to page-charge-based cost-recovery and free distribution in place of S/SL/PPV will mean that journals -- competing for authors' papers rather than readers' payments -- are free from the need for firewalls to segregate papers from readers, and can instead collaborate in launching scholarly skywriting into the PostGutenberg Galaxy.