Typesetting is a huge bottleneck in scholarly publishing. What can we do to make it easier?

Great article, and one that applies far beyond academia.  I'm a web designer and programmer and receive content for publication on website daily, usually in MSWord format.  Approximately 10% of my job is converting MSWord to plaintext and then marking it up in HTML to fit, something which (frustratingly) my clients seem happier to pay for than to simply use notepad and very simple tags, nothing more complex than an HTML anchor. 

There's a very clear split between people who understand HTML and are willing to separate content and style in their heads, and those who feel that the font is as important, if not more so, than the informational content of the document concerned.  Some have even demanded to know why their website looks different on their iPhone and their Windows XP desktop.  Showing them what it looks like on a text based browser like Lynx ("this is how Google sees your website") is usually a shock to them. 

I think the solution to the problem is at the top of the editing window I'm typing into.  There's a series of buttons that let me do thing like bold, italics and

• even
• bullet
• points

That's basically MSWord, yes?

OK, most people reading this will understand that it's a JavaScript browser plugin and very different to a standalone application, but to the people who don't understand why HTML is a good idea, it's basically MSWord.  They click buttons, their text looks pretty.  That's more important to them than semantic information and machine-readable structure, they have people to do that for them, the journal publishers.  The good thing about this system, however, is that the output is relatively clean HTML.

It's relatively simple to put together a browser based application that they can run locally, just like Word. It'll look a little different, but the difference is far less than asking them to learn a "programming language", which is how they think of HTML.

Basically, I'm suggesting a new word processing standard for academic publishing, based on existing, cross-platform and open source technology that uses a familiar GUI but outputs good, clean, easily edited HTML.  Just what it was meant to be back in 1990 in fact.

If journals started offering a BIG discount on manuscripts written in something other than MS Word, I'd consider it.  BIG means ≥$500.  Not that I like Word (I hate many aspects of it), but it's what I know and what the people I interact with are comfortable with.

Otherwise, rather than asking that many thousands of researchers each take the time to learn a new way of preparing their manuscripts, and to persuade all their collaborators to do likewise so they can all read the drafts, why can't someone with the necessary expertise write a program that converts MS Word docs into whatever works well for final typesetting.  They could sell it to the typesetters/publishers for some of the moey that would otherwise be used to provide the BIG discounts.