This is excellent news. The biggest hurdle for any group of researchers getting together and doing a journal The Right Way seems to be the notion of perceived prestige that still clings to Scrooge-journals. Go Gowers!
It seems to be all about the moderation system. I wonder if there are any books about moderation systems, describing theory (how to make sure the system cannot be gamed, how to rank entries, how to rank users, how to ensure decidability etc.)
He's deliberately making it extremely conventional in every way except where the papers are hosted. At this point the game is to gain prestige. Elsevier et al will have a hard time knocking this one down.
Elsevier et al have so far played very nice with arXiv, for instance by not demanding that papers submitted to Elsevier journals not appear on it.
Technically the arxiv may right now be illegally hosting many thousands of papers whose authors actually ticked a box granting exclusive publication rights to Elsevier. Or any number of other journals. Elsevier benignly neglects this, but if you really start threatening their business model then they can make life very difficult for the arXiv.
Basically Elsevier supporting the Research Works Act was the last straw. (And what inspired Gowers to start the boycott.)
They could do this, but it would be pretty much the final fuck-you. Not that that makes it implausible, of course.
Look at the backlash at Wikipedia for just accepting account access for editors from Elsevier. That sort of arrangement has gone on with various academic publishers for years - but in 2015, it creates a backlash. (I can assure you that in and around Wikimedia, pretty much everyone's actual opinion is "Elsevier delenda est".)
One intrinsic drawback in the concept of overlay journals is however that is does not allow for double-blind reviews.
Until this is widely implemented I think the publishing process won't be fair.
I'd be fine with reviews being openly attributed to the author. Authors reviews could be counted as contributions. Also, it might improve review quality.
Scholastica's practice of charging $10 for each submission seems ridiculous: if someone submits a million dummy submissions the journal has to ask support to forgive the charges...
I think that a much better approach would be to require papers to be submitted as Git pulls adding TeX files, and then use a free code review tool like Gerrit to do peer review.
That might take a bit more learning than Scholastica, so starting with Scholastica until the journal is more popular might make sense, but using such expensive proprietary software doesn't seem to be acceptable long term, and if a tool works for peer review of code changes, it should work for peer review of articles as well (maybe with some customization if needed).
Ten dollars is nothing. As others have noted publication charges tend to run more in the thousands.
And talking of programming tools, like git, and gerrit for academic papers is completely off base. Most mathematicians will never even have heard of git.
The "free" version is something like what the episcience project is attempting.
Edit: And as yet others have said, it's crucial, especially in a structurally very conservative field like mathematics, to stay as close to the conventional model as possible.
> but using such expensive proprietary software doesn't seem to be acceptable long term
Is $10 per paper really that expensive? The recent papers I've published have come with page charges in excess of $1000 (in a not-for-profit journal). Obviously it's far from an apples–apples comparison, because the $1000 pays for things like the journal having a copyeditor to go through the manuscript and also results in reduced subscription fees (~5x less than another comparable journal in the field). Though, for this "overlay" journal, the subscription point is obviously moot. But, from an author's perspective, $10 seems like a steal.
This makes it impossible to do blind review (review without knowing who the author was), which seems desirable. Do existing math journals do blind review?
Blind (or double-blind) review in most of academia is inherently impossible -- the people qualified to do the review will generally be able to recognize who the paper must have come from.
I agree that blinding would be preferable, but when the material itself reveals the author it's simply not possible.
None of the journals I've submitted to do double-blind review.
Usually people post their preprints to the arxiv at the same time as sending the article to the journals. This means the information (in a somewhat rougher form) is available immediately, and priority on an idea is established quickly. This has greatly sped up the development of mathematical ideas, but, yeah, essentially means one can search for the article being reviewed, and that double-blind review is more-or-less impossible.
OTOH, double-blind review is a great way to deal with conscious and unconscious institutional and gender biases, so would be great to have...
