Journals for certification,
The PUBlicatiOn cUltUre in computer science is different from that of all other disci- plines. Whereas other disci- plinesfocusonjournal publication, the standard practice in CS has
been to publish in a conference and
then (sometimes) publish a journal version of the conference paper. Indeed,
it is through publication in selective,
leading conferences that the quality of
CS research is typically assessed.
Why should a researcher publish
a journal version of a paper? In fields
other than CS, which place no value on
conference publication, there are two
˲ ˲ Certification. Publication in a peer-reviewed journal is a signal to the world
that the paper has passed a minimal level of competence; publication in a leading journal confers even more prestige.
˲ ˲ Publicity. Journal publication can
be an effective way to tell the world
(including policymakers and science
advisors, not just colleagues) about the
research, particularly publication in a
leading journal like Science or Nature.
In CS, the situation is different. In
many subdisciplines of CS, having a
paper accepted at a leading conference
already gives as much of a certification
as getting it into a leading journal.
Conferences play the role of publicity
as well. The best way to get your subdiscipline to know about your results
is to publish them in the leading con-
We can already
see the beginning
of a shift in the
ference for that subdiscipline.
There is increasing debate about
the role of conferences in our field.
1, 5, 8
Fortnow3 argues that our field suffers
from the current use of conferences
for certification.a Two particular problems he cites are those of quality and
innovation suffering because we end
up living in a deadline-driven world,
and the splintering of the field into
multiple conferences (so that there
are enough publication venues), leading to conferences failing to act as a
broad forum and bring their communities together. Fortnow suggests conferences should be held less frequently, and accept every reasonable paper
for presentation without proceedings.
a We remark that Fortnow’s Viewpoint is not
universally accepted; see, for example, the
different reactions noted in in Vardi’s March
2010 Communications’ Editor’s Letter.
Even in the current situation, journals do play a role in providing more relaxed page limits, which allow authors
to include more discussion, more
expository details, details of proofs,
additional experimental results, and
the time to submit a more polished,
thoughtful paper. For theoretical papers, the certification issue remains
significant because it is rare that conference reviewers review proofs as
thoroughly as journal reviewers. Publication in a journal also adds value
through a paper going through a strict
review process with several iterations.
Finally, journals also provide publicity
for interdisciplinary work.
We can already see the beginning
of a shift in the conference and journal landscape. Part of the shift involves journals publishing conference
proceedings as special issues. For example, ACM Transactions on Graphics
( TOG) publishes every SIGGRAPH and
SIGGRAPH Asia technical paper in its
biannual conference issues, which replace traditional conference proceedings. If a paper is conditionally accepted for presentation at SIGGRAPH,
then the paper undergoes a second
review (by one of the original reviewers) to ensure all changes requested by
the reviewers are made; it is then also
published in TOG. Similarly, papers accepted at this year’s International Conference on Logic Programming (ICLP)
will appear in Theory and Practice of