Logic Programming. However, note that this approach means these journal issues suffer from the same problems that conference publications suffer from: papers are subject to the conference length restrictions and paper submission deadlines that make it difficult to do serious revisions.

The database community has taken this one step farther. Currently the only way to submit a paper to the VLDB (Very Large Database) conference is to submit it to the journal Proceedings of the VLDB Endowment (PVLDB). Continuous submissions are accepted throughout the year, reviews guaranteed within two months, a full review cycle including checking of final versions by responsible editors is supported, and papers accepted by a specified date are offered a presentation slot in the next VLDB conference. This change makes it possible to have a “ revise and resubmit review,” but there still remain serious page limitations. (Interestingly, PVLDB also allows the publication of an extended version of a PVLDB paper in another journal.)

The TOG/SIGGRAPH relationship has another facet, which illustrates another possibility for a conference-journal relationship: TOG allows any author of a published paper to present the work at SIGGRAPH, while operating without deadlines and less strict page limits.

IllustratIoN by stuart brIers

These experiments suggest that the
CS community needs to think through
the intertwined role of conferences
and journals, especially in light of the
growing amount of research at the in-
tersection of computer science and
other fields. Our unique distinction
of being a conference-led field leads
to a particular problem for multidis-
ciplinary work because, outside of CS,
journals typically have all the power,
and are very reluctant to take papers
where versions have appeared previ-
ously. For example, some leading biol-
ogy journals are unwilling to publish
work that has appeared in RECOMB
(Conference on Research in Compu-
tational Molecular Biology) or ISMB
(Conference on Intelligent Systems for
Molecular Biology); this has caused a
problem for researchers in computa-
tional biology. Some interdisciplinary
conferences are sensitive to this is-
sue (including ACM Electronic Com-
merce), and allow full papers to be
submitted and reviewed, but then pub-
lished as a one-page abstract in the pro-
ceedings. However, this is not the norm.

future Role of Journals

We offer the following vision of the future role of journals within CS, with some thoughts on how to make it come about. Many of these ideas have been suggested and, indeed, some have even been tried. But more serious experimentation is needed. Our vision of the future includes the following elements.

Papers will be available on public Web archives such as CoRR, the Computing Research Repository (see http:// arxiv.org/corr), the CS part of the arXiv. This is increasingly common now, as researchers are discovering the advantages of posting papers on managed archives rather than just having them on their own home pages. But if all papers are available in one place, then making a paper stand out from the pack will become more significant. One of the best ways of doing this will be via certification.

Journals will be the main “ certification” authorities, because they can operate without deadlines or strict page limits, allow for a careful review cycle with checking of results, are compatible with other scientific disciplines, and promote thoughtful work. (By “journal” we simply mean an editor-in-chief together with an editorial board recognized by the community as a certifying authority. Journals like Journal of Artificial Intelligence Research (JAIR) and Logical Methods in Computer Science (LMCS) demonstrate that a group of community members can start viable, well-respected journals without the support of a publisher.) To paraphrase Churchill, we believe that journal reviews are the worst form of certification, except for all the other ones that have been proposed now that papers are available online. For example, citation counts (and page rank style variants of citation counts) suffer from well-known problems, including the fact that different fields have different citation rates, they can be influenced by fads,