Communications of the ACM

David Kaeli

The top quality conferences in my field are either co-sponsored by ACM and IEEE-CS, or are sponsored by only one of these organizations. Would it be possible to have ACM and IEEE-CS work together on the ambitious idea put forth by ACM, but to be inclusive and carry the endorsements of both organizations? I know it would benefit my research community!

Christian Cachin

It is great to see such a proposal and discussion at the ACM level. I sympathize with its goals but am missing some crucial information.

As PACM would "publish the proceedings of ACM conferences", would its reviewers fully replace the conference program committees? If yes, the criteria in the bullets would simply be conditions on the review board (alias program committee - PC) of a conference. This would change some of the nature of the PC work, which McKinley's viewpoint heralds.

Would the current operation of PVLDB (Proceedings of the VLDB Endowment), with its continuous submission option, be compatible with PACM?

Much of the conference culture revolves around the perceived prestige or competitiveness of a venue. If multiple conferences are part of the same PACM for XYZ, would a distinction of which conference selected the paper be made and encouraged, or would this be hidden? Would there be special issues by conference venue so that authors can still cite their work referring to the conference?

Joseph Konstan

Thanks for your comments (and we'll keep trying to respond).

David, the idea of ACM and IEEE-CS working together is intriguing -- right now both organizations are experimenting with different ideas, but if we hit the right thing for the field, it is clear that we (and perhaps AAAI, SIAM, and many others) should coordinate.

Christian, the intent is not that PACM has reviewers, but that the reviews are done by conference program committees (and then the criteria become conditions on the review process). For some conferences, this would be a major change; for some, no change at all. And, of course, we're still seeking input on what the important factors are for those standards. The question of PVLDB is an interesting one (that motivated some of this exploration). One could argue that it fits this model, or that it fits the "journal-first" model (albeit with a journal whose model doesn't quite fit the ACM Transactions approach). But the concept of year-round submission is one that the task force was certainly in support of. Indeed, there are no rules preventing today's ACM conferences from moving to that model, though the logistics are different and potentially challenging. As for the individual conference proceedings within a series, the idea is to have an issue for each one, but it is also to ensure that the overall quality of a series is maintained at a high-prestige level. The task force's hope (and that of many commenters) is that peer pressure would lead to maintaining high standards while at the same time encouraging SIGs and conferences to open up venues for less thoroughly reviewed work, such as more workshops.

Robert Harper

In my view it is simply not the case that conference papers are equivalent to journal papers in other disciplines. Nor is it the case that they ought to be. Conferences satisfy a timeliness requirement; journals satisfy a reliability requirement. Results presented at conferences should be regarded as tentative until a backing journal article is published. Conference papers are not and should not be considered archival. If a tenure committee wishes not to consider conference papers as suitable evidence of scholarship, that is a rational and reasonable decision on its part (as would be the opposite decision). It is wrong to try to jigger the system so that conference publications look like journal publications. You've really got the whole thing backwards!

Amir ben-Amram

Many ACM conference papers are already quite long; a 12-page two column POPL paper
that I refomatted to TOPLAS format filled as many as 25 pages. And PC members have
their hands full; some PCs near or exceed 40 papers per PC member.
My opinion is that removing page limits from conference publications is not going to
help with the difficulties of conference reviewing and the quality of published papers.
Such an experiment was actually conducted at the FOCS conference series. The experiment has been abandoned by FOCS 2015 and the page limit reinstated.

The PACM proposal is somewhat ambiguous in this respect. It suggests to remove page
limits but also allows for non-refereed supplementary material, an option which allows
for the common practice of submitting appendices together with the conference submission.
It seems clear to me that the proposal does not try to change the main characteristics of
conferences. Reviewing will still be done by PCs (and perhaps their ERC or subreviewers)
and subject to the customary deadlines (more or less).
Also the quantity of conference papers is not expected to
change (unless it grows to accommodate the demand for publication in PACM).

I conclude that PACM papers are conference papers and should not be seen as the same
kind of entity as traditional journal papers, which undergo a more thorough, slower
refereeing cycle.
I believe that for many papers, this is not a problem. For example, some conference
papers may represent steps in a big project which might usefully culminate in a single
journal paper. Other papers are time-dependent. Some conference papers may actually be
self-contained, within page limits.

But there are also papers that cannot serve their scientific purpose without the long
(and sometimes hard) process of traditional journals. Papers dependent on subtle
mathematics require careful refereeing. With such papers, a conference paper - undergoing
traditional conference reviewing - has to be considered a preliminary announcement of
results (traditionally, such papers were often referenced as "extended abstracts").
I am wondering how the PACM model can accommodate this kind of papers.
Perhaps the two-track idea now adopted by ICLP should be considered - see
http://booleconferences.ucc.ie/iclp2015. Note that they allow both the author and the PC
to put a submission into the "non-journalized" track. This may alleviate the conflict
of the PC when presented with a paper that may leave some things to be desired (perhaps
even after a single revision as permitted by the PACM proposal), or that they feel they
cannot check in detail, while they also feel that the paper is significant and its
presentation to the public should not be delayed.