I'm not a publisher but have worked at publishers in the past.

> All the journal does is provide a crappy web app > to organize the process.

I'm guessing you're being facetious ;) but that's not true. It depends on the publisher but they may also edit the manuscript; find the right reviewers; have a full time editor make a decision based on the reviewer's feedback; typeset or convert it to machine readable format (authors submit in Word, but this'll often be converted to an industry standard XML format to allow interoperability between different databases - both abstract databases and publisher ones as journals change publishers frequently); ensure the content is properly stored (this is not the same as putting it up on S3 for 2 cents a month. It's the scholarly record and publishers generally take the responsibility of maintaining it seriously. Publishers pay a fee to put content into a darknet run by a consortia of libraries - if the publisher goes out of business the consortia makes all the content available for free. They also pay membership fees to industry organizations like CrossRef that allow them to deposit canonical metadata in a central database and mint DOIs) and discoverable (at the end of the day your paper will get much more attention when published in Science than it does your blog).

Are all of those things worth paying more than $99 an article for? It's debatable.

But can anybody with a Wordpress blog do what publishers can? No.

(Same sort of thing applies to trade. Look at self-published books. You can create great content for free and stick it in an EPUB but you'll rapidly find that you need to start shelling out to be included in the book databases bricks and mortar stores use, to have a cover designed, to get an ISBN, to start actually selling through Amazon or wherever... publishing isn't as trivial as it seems.)

I'm only familiar with math publishing, but I've never heard of a math journal that did anything other than provide a web form for the editor to request a review (and similarly, for the reviewer to submit with).

Typesetting is usually done (correctly) by latex, and then possibly screwed up by some human the publisher outsourced to. Automated typesetting is a solved problem, and there is no need for humans.

As for your cost of $99/article, the Arxiv costs $400k/year (though admittedly, Donald Knuth does the typesetting).

As for your "darknet", are you really going to claim this darknet is more reliable than mirroring articles to 3-4 amazon regions plus rackspace files? I'm sure the darknet costs at least 10x as much, but that isn't the issue. Similarly, paying membership fees to the other big publishers to keep little guys out of the market is again not the issue.

If it weren't for reputation effects and agency costs (i.e., if the market were competitive), this process would cost well under $99/article. The only reason the publishers haven't already died is because tenure committees want to see Journal of Computational Physics rather than the Upstart Computational Physics Journal.

Don't get me wrong - I don't think that the existing system works particularly well or is value for money, I just wanted to point out that there's much more going on than hosting a crappy website. ;)

> only familiar with math publishing

Fair enough. I'm only really familiar with the life sciences.

> Automated typesetting is a solved problem

This isn't as simple outside of CS/EE/maths unfortunately. Most academic researchers couldn't tell you what LaTeX is.

> As for your cost of $99/article, the Arxiv costs $400k/year

arXiv is awesome but it's a preprint repository not a journal (whether or not a repo + engaged community is as 'good' as a peer reviewed journal is probably a whole other thread... physicists still usually publish in journals, even though they use arXiv constantly and put their manuscripts up there first)

I picked $99 because it's the number PeerJ (also awesome) is suggesting. In reality life science journals will charge you $1k+...

> are you really going to claim this darknet is more reliable than mirroring articles

I don't make that claim, academic librarians and publishers do.

(if you prefer you can swap out this cost and replace it with the cost of hiring somebody to do ops for you on an occasional, freelance basis to handle things going wrong, moving to new cloud providers etc. over a 25 year period...).

> Similarly, paying membership fees to the other big publishers to keep little guys out of the market

I don't think I explained this the right way. The fees are to help maintain common services that all publishers then use - for example, to ensure that given an identifier you can always find the full text of a paper (URLs being unstable over even relatively short timeframes), to mint and store metadata about those identifiers to make sure the right people get credit when cited etc. Those services are unquestionably valuable and get used by millions of scientists every day.

> If it weren't for reputation effects and agency costs (i.e., if the market were competitive), this process would cost well under $99/article

I disagree but believe that you should be right. People like PeerJ, eLife, arXiv and others need to keep exploring alternative ways of doing things until we hit on a sustainable model.

> (if you prefer you can swap out this cost and replace it with the cost of hiring somebody to do ops for you on an occasional, freelance basis to handle things going wrong, moving to new cloud providers etc. over a 25 year period...).

Plenty of countries have national/legal deposit libraries (my country has six of them) and the costs of running them are a rounding error in national terms. You can hardly claim an electronic national library would cost more to run than a physical library - much smaller volume of stuff to store, no staff to retrieve and shelve documents, no in person visitors.

I take it you haven't submitted to many CS journals ;) Last time I made a submission to the ACM I had to essentially hand in a publication ready pdf, this is true for the vast majority of CS journals. Additionally I've published in open access journals outside of CS where I did just have to submit a word doc and they took care of the rest.

Additionally I've known plenty of editors who do many of the things you mentioned above for free (for paid journals) with the only reward being bonus points for tenure.

Considering that there the costs to creating software are much higher (simply because unlike journal articles there's no real incentive for devs to give away there rather price labor, other than just wanting to give back) and we still have a thriving OSS eco-system I find it very hard to believe that we can't easy substitute out the very few parts of academic journals that do incur some sort of real cost for the publisher with volunteer work.

"Publishers pay a fee to put content into a darknet run by a consortia of libraries - if the publisher goes out of business the consortia makes all the content available for free"

So driving them into ruin collectively frees the back catalog? Since access to old papers are a major concern in every plan to cut the middlemen, that would actually be good news!