Submitting ACM Conference papers to ArXiv

Question

I'm a little confused. The ACM Author Rights and Publishing Policy states:

ACM authors hold the right to post copies of their own peer-reviewed accepted ACM works on any non-commercial repository or aggregation that does not duplicate ACM tables of contents, i.e., whose patterns of links do not substantially duplicate an ACM-copyrighted volume or issue. Non-commercial repositories are here understood as repositories owned by non-profit organizations that do not charge a fee for accessing deposited articles and that do not sell advertising or otherwise profit from serving articles.

However, the ACM Fair Access site states:

Can I publish the published version of my publication on arXiv? No. An ACM author cannot grant arXiv.org or any other publisher rights to distribute their works that have been published by ACM.

SHERPA/Romeo shows ACM as being a "Green Publisher", which would support the case that I can upload to ArXiv.

Dissem.in states that I can upload pre- or postprints under certain circumstances, but does not explicitly mention Arxiv, positively or negatively.

So, what is it? Can I submit a paper I published with an ACM Conference (ACM WiSec, if that makes a difference) to ArXiv or not? As I understand it, ArXiv is a non-profit, so it would fit the terms of the publishing policy. Or is this related to the rights you have to grant ArXiv upon submission?

If I can only publish a specific version of the paper, which one?

Answer 1

This is a matter of obsolete documentation.

The current version of the "ACM Author Rights and Publishing Policy" you linked to is v9, last revised in 1/2016. This version says that authors have the right to:

Post the Accepted Version of the Work on (1) the Author's home page, (2) the Owner's institutional repository, (3) any repository legally mandated by an agency funding the research on which the Work is based, and (4) any non-commercial repository or aggregation that does not duplicate ACM tables of contents, i.e., whose patterns of links do not substantially duplicate an ACM-copyrighted volume or issue. Non-commercial repositories are here understood as repositories owned by non-profit organizations that do not charge a fee for accessing deposited articles and that do not sell advertising or otherwise profit from serving articles.

(emphasis mine.)

The previous version, v9 with a 5/2015 last revision date, only gave authors the right to

Post the Accepted Version of the Work on (1) the Author's home page, (2) the Owner's institutional repository, or (3) any repository legally mandated by an agency funding the research on which the Work is based.

The Fair Access FAQ page is from before January 2016, and reflects the old policy.

The current policy as of January 2016 does allow the accepted version of the work to be posted on "any non-commercial repository or aggregation that does not duplicate ACM tables of contents" such as arXiv.

malexmave confirms after contacting ACM:

FAQ was indeed outdated. Posting the peer-reviewed / accepted version to ArXiv and other non-commercial repositories is fine (but not the final version). Policy change does not apply retroactively, only to new papers under the new policy.

Answer 2

The ACM offers guidelines for Prior Publication and Simultaneous Submissions. Information there (from April 2018) extends this answer if the paper is a technical report prior to submission or under review as opposed to being already accepted. The guidelines state

Issuing the paper as a technical report, posting the paper on a web site, or presenting the paper at a workshop or conference that does not publish formally reviewed proceedings does not disqualify it from appearing in an ACM publication.

I assume for the case that the manuscript is not yet submitted or still under review, arXiv would count as "issuing the paper as a technical report, posting the paper on a web site."

I understand the following: Overall, ACM usually allows researchers to archive their manuscripts prior to submission, during review, and after peer-review/acceptance as long as the manuscript is not equal to the official ACM publication and, after publication, contains a corresponding link to this official (i.e. formally reviewed) publication.

Please, let me know if this is incorrect. Some philosophical background on why publishers should be less restrictive anyway.