New York, NY Thursday, January 1, 2015 -

For the annual CR editorial, we’ve invited Koen de Bosschere of Ghent University to contribute his views on the possible evolution of conferences and journals as CS publication venues. He has led and observed positive change, both within the journal he edits and its most relevant conference.

The publication culture in computing is becoming more diverse. An increasing number of authors (mostly outside of the US) are expected to publish papers in journals in order to get tenure or promotion. The dominant publication culture in computing, however, is conference publication, and journals are often used to publish extended versions of papers previously published at conferences. Since most authors prefer to cite the original (conference) publication, the extended journal versions are often cited less, leading to relatively low journal impact factors for CS journals. This situation is not serving well the authors needing journal publications for advancing their careers.

Some computing journals therefore recently started to change their review processes: they call for original work manuscripts and are promising short decision deadlines (one-to-two months), and immediate digital publication after acceptance. As a result, some of these journals already publish papers faster than major conferences.

In the longer term, this might lead to two parallel communities, each with its own publication ecosystem: one that needs publications for a limited list of very selective conferences, and another one that needs publications for a list of high-impact journals. In practice, authors might be forced to choose one of the two models. Spending time on the alternative model is a waste of time because it does not advance one’s career.

The effect might be that mobility between the two communities becomes increasingly difficult. I am aware of at least two cases in which excellent US candidates for professorship positions in Europe did not make it because they did not have a good enough journal publication record (including journal citations, reviewing for top journals, and so on). I can imagine that candidates with a good journal publication record, but without papers in the best computing conferences, will not impress US hiring committees either.

I believe that maintaining two publication models will eventually divide the community. Authors will always focus on the model that helps advance their careers. By supporting the two communities, ACM will unwittingly push the two communities apart, which would be a bad thing.

One conference (HiPEAC) has found a solution to temporarily meet the needs of the two communities. In 2011, it decided to collaborate with ACM Transactions on Architecture and Code Optimization (TACO) to take care of its complete paper selection process. All papers submitted to the HiPEAC conference are forwarded to ACM TACO, which does two full review cycles in five months, and immediately publishes the accepted papers in the ACM Digital Library. The authors of accepted papers are invited to present their work at the conference (90% accept the invitation). This model has led to a fourfold increase in the number of submissions to ACM TACO, and has at the same time doubled the attendance at the conference, which has since become a true networking event for the community. This model, which is known as the “journal-first publication” model, tries to combine the best of two worlds: a genuine journal publication and a presentation at a major conference.

Koen De Bosschere

Ghent University

Interim Editor in Chief, ACM Transactions on Architecture and Code Optimization

http://taco.acm.org