In physics I doubt if this is an issue, as most important journals are published by professional societies.

Everyone could shove whatever they liked into an ArXiv-like repository, but the only journals that would survive would be those like Nature or Science.

Well, but the current state of affairs seems to contradict this argument. Everyone can “shove whatever” on ArXiv, but some preprints are read (and cited — I am sorry but I need to insist on this aspect), other ignored. And we have many more journals in physics than just Science and Nature. Who knows, maybe my model would work better for physics than other disciplines but I don’t see what is fundamentally different about physics.

consigning science publishing to the cut-and-thrust of market forces is not where I would like to see it go.

… which is why I advocate the diametrically opposite approach of regarding as “published” anything that is uploaded on the archives, and let the number of citations be the real metrics to asses the paper’s worthiness – which is, incidentally, exactly what we are already doing anyway (I’ll take an unpublished preprint cited 100 times any day over a PRL that is never cited), except that we have to go through a long, time-consuming, expensive, aggravating and ultimately useless process of “vetting” by journal publication.
In my model, journal publication would be essentially only a somewhat more official, a posteriori recognition of the validity of the work (akin to its inclusion in the “Focus” section of Physical Review Letters, for example), but no more than that. Again, citations are what matters in the end, not where the paper is published.

I have colleagues that edit relatively small journals, put out by their professional society, that survive because of subscription revenue, but I’m sure wouldn’t if they were supported by ad-revenue. These play an important part in the ecosystem of science and it would be a travesty if they were put down.

I have a hard-time seeing how ad-revenue only (or bulk) wouldn’t evolve into a two-tier system. Everyone could shove whatever they liked into an ArXiv-like repository, but the only journals that would survive would be those like Nature or Science. While that landscape for physics might be acceptable, page-views and click-throughs will be fundamentally different between disciplines, simply by the nature of how they use the literature (analagous to the discipline-specific distributions of h-indices). Some otherwise worthy journals in disciplines such as mathematics and paleontology would just die.

If one is willing to treat “journals” like glossy magazines -consumable and of no real consequence- that’s fine, but otherwise consigning science publishing to the cut-and-thrust of market forces is not where I would like to see it go.

Anyways, your proposal is very though provoking, thanks for bringing it up. I’m not completely convinced my arguments are correct, but surely they can’t be too far off(?).

My friend, I have got news for you: that is the way it already is. In my model at least you have the paper on ArXiv and it can still be cited, if no one journal will pick it up — and it if gets cited all the time, eventually some journal will publish it.
Maybe there is no room for all the, say, physics journals that exist nowadays, but I am afraid that they are doomed no matter what we do.

No, advertising should instead be the bulk of a journal’s revenue, in order to keep publication charges down.

Obviously that would be a bit extreme, perhaps a system where advertising was only a (small?) part of the revenue stream would work ok? Or some sort of revenue sharing?