Savings from Converting to On-Line-Only: 30%- or 70%+ ?

From: Mark Doyle <doyle_at_APS.ORG>
Date: Wed, 26 Aug 1998 22:22:56 -0400

Hi,

The American Physical Society costs for journal production are
somewhere in the neighborhood of 1/3 for editorial/refereeing, 1/3 for
composition and copy-editing, and 1/3 for printing and
distribution. So I think your 70/30 split is quite reasonable.

The problem with some other analyses is that, as you point out, they
don't even begin to analyze how the proper use of electronic media can
greatly reduce the costs associated with the first 70% -- it is clear
that to see such dramatic savings, it is necessary to transform the
entire process, not just the final distribution. However, this
transition requires the participation of the authors and referees,
which is a major impediment to a fast, smooth, and low-cost
transformation.

xxx solves quite nicely the problem of getting submissions into the
publication process, out to the referee, and distributing the work to
readers. The APS now makes good use of xxx for the first; Phys. Rev. D
(the high energy physics piece of Physical Review) makes some use of
the second (other more conservative fields are reluctant to use xxx
and would rather have an APS controlled referee server); and the APS is
reluctant to do the third (i.e, make xxx the primary means of
distributing Phys. Rev. papers). However, as usage of xxx grows and
higher percentages of papers can be found on xxx, there will be a lot
of pressure on the APS to make the content of papers freely available.

The real danger is that subscriptions to our journals could dry up
before we have a chance to complete the transition to a new, leaner
all-electronic publishing process. The APS (a non-profit publisher
with benevolent aspirations) is much more vulnerable to this than
prosperous commercial publishers with deep pockets. Luckily, we have
the prestige of Phys. Rev. to sustain us in the short term.

In any case, the thinking around here is in terms of selling services
wrapped around the articles, not necessarily the articles themselves
(linking to other articles, linking to other resources, full-text and
fielded searching, etc.). This of course would mesh quite well with just
overlaying xxx, but the marginal cost of just serving up the articles is
quite small in comparison with the rest of the services...

The major shortcoming of xxx when compared to what the APS does with
articles is in the uniformization of information into standard, long-term
archival formats, namely, fully tagged SGML, and the maintaining of an
extensive database used for referee selection. Currently the APS (and I
suppose all print publishers) handle this by throwing large amounts of labor
at the problem. Manuscripts are converted where possible, but many are
still re-keyed. AIP (our main vendor for composition/production) really
hasn't yet sorted out what they could be doing in this area.

We would like to have all of our manuscripts prepared in a way that would
cut out most of the labor -- for instance, all information in the
frontmatter and bibliographies should go right into our databases with no
manual intervention and then this should be used to directly feed the
referee and production processes. To do this requires authors to be
disciplined and to consistently use high-level TeX macros (new REVTeX
for instance) or a new Word template we have been developing. It
remains to be seen how authors will take to such tools.

Note that whatever aids the APS in automated conversion can be immediately
adopted by xxx -- thus we won't be at any advantage; we will only have
been able to bring our costs more in line with what is optimal.

The other problem we face is that the majority of our authors and referees
are foreign and many still can't cope with the web or even e-mail. We still
rely on faxing and snail mail for a large part of our communications with
authors and referees and it will take a fair amount of time to reduce this
(though it is inevitable that this will improve).

The APS did start a new journal Phys. Rev. Special Topics in Accelerators
and Beams. This journal is 100% electronic - the paper process that is used
to track manuscripts has been made over into an electronic one and all
referrals and communications are electronic. Also, no print version exists.
As it grows (and as we migrate the applications prototyped here to our other
journals), I think we will have a better gauge of how the costs compare
between electronic and paper publishing. Also, it should be noted that this
journal is available to all for free because of a new economic model in
which a consortium of interested parties (like SLAC) has put up money to
help defray the costs. This model is in addition to those in your
S/SL/PPV troika.

About preservability, I would agree that upgrading is straightforward
to the extent that PostScript can go to PDF can go to successor of
PDF... Information won't be lost. However, it should be noted that in
the conversion process, it isn't always possible to take full advantage
of the new features in the new format unless one has a decently
marked-up version of the original source. Hence, the successful use of
HyperTeX at LANL in many old papers in producing richly linked PDF
files, but the failure of many others that are just PostScript papers
that can't be promoted to hyperlinked PDF's. The APS spends a fair
amount of money creating the fully marked up version of articles so
that they can be used later to create new formats (well, that is in
principle: in practise the process is much more costly than it should
be because the SGML is produced as an afterthought, rather than as the
central deliverable).

A real problem is that authors are extremely slow to adopt better tools
that make electronic publishing more cost-effective. Hence large parts
of the physics community insist on using Word or poorly marked up TeX
files. Only large, established print journals are set up to apply
manual labor to make up for the shortfall on the author side.

Good authoring tools do not exist at the moment and this really is a
large part of the current costs. It affects many things (for instance,
it would be too costly to convert all manuscripts early to SGML (thus
saving several weeks in the publication process making the journals
more useful for dissemination of new results) and it makes it hard to
do author corrections because authors can't work with the SGML (or our
other internal composition systems -- proof changes require manual
editing). And only when authors systematically tag their documents
early will it be possible to optimize the building of citation and
other linked databases.

Disciplined use of good authoring tools that promote good mark-up
would play a major role in bringing down the current "paper
cardhouse", making possible a substanital reduction in the labor costs
now associated with producing any large journal. Otherwise, I think
some communities will really lose something by such a collapse.

Many manuscripts are currently submitted to the APS via xxx through a web
form interface. And Phys. Rev. D does tell referees that they can use a
particular version on xxx. But I am afraid that this practice won't spread
to our other journals and the move is towards the APS having its own
referee server (though I think there will still be pressure in the PRD
community to stick with xxx since it is so well-adapted to that
community). Still, all APS journals will link to xxx for e-print
references (and we expect xxx to link back to us).

Regarding copyright: The APS explicitly recognizes the rights of
authors to circulate their work through xxx, even after APS
certification (and even the revised manuscript can be circulated in
this way).

Cheers,
Mark Doyle
APS Research and Development
doyle_at_aps.org
Received on Tue Aug 25 1998 - 19:17:43 BST