Semantic TeX - where are we heading?

Question

(Update: I learned that this kind of open question is not supposed to be asked here at tex.stackexchange.com, so I will refrain from furthering the discussion here and invite contributions instead via the link provided in the next line.)

Overheard today:

I believe what is likely to happen is that some alternative language will be developed which is syntactically similar to TeX (since this will make it easier for mathematicians and physicists to transition to it) with less focus on typesetting but far richer semantically, allowing a smoother transition to MathML.

A brief Google search revealed that a number of initiatives have been started along these lines, e.g.:

• SALT (Semantically Annotated LaTeX)
• Hermes
• STeX
• SnuggleTeX

My question, now, is how to get an overview what the comparative strengths and weaknesses of these (and other) approaches are, what communities and applications drive them, and what else to consider. I would be surprised if there were a simple answer to any of these, so I expect to split this thread up into simpler ones as the discussion develops.

Answer 1

I would classify gellmu http://www.albany.edu/~hammond/gellmu/ as well as possibly the XML layer of plastex in this category as well.

First of all, I think we should distinguish two meanings of "semantic" in this discussion. There is the old fashioned sense in which HTML is semantic, i.e. it specifies content rather than how it is to be presented, but there is also the "web3.0" concept of semantic in which you have richer ways of specifying things like the relationships between items of data. Since LaTeX is not even semantic in the first sense, I think it is far more important to discuss that notion of semantic than the more modern one.

For me the main interest in these things is to enable mathematical scientists to make use of rich web tools for collaboration, authoring and presentation of mathematically oriented work, e.g. specialist tools for things like the polymath projects. This would indicate that XML or HTML based formats should be the way forward. However, for the foreseeable future, it seems likely that mathematically oriented journals will continue to use LaTeX as their primary submission format. It also seems likely that academics will continue to publish peer reviewed manuscripts as the main way of securing credit for their work. These things may change in the long term, and arguably they should, but that is the reality that we are dealing with right now. For these reasons, any workflow that is likely to catch on must involve a markup language that is similar to LaTeX (since academics have to learn that anyway), must be capable of being converted both to high quality LaTeX and XML/HTML, and must not suffer too badly from conversion from LaTeX back to the chosen authoring format, e.g. this may be needed to make corrections in the peer review process.

There are various reasons why LaTeX to HTML converters cannot be expected to behave perfectly. For one thing, LaTeX inherits many typesetting specific constructs from TeX that have no meaning in the HTML world. For another thing, some people are very fond of pointing out that LaTeX is Turing complete, whereas HTML is not. For these reasons, it is useful to identify a semantic core of LaTeX that can be converted to HTML/XML reversibly, i.e. things like \section{XXX} would be fine, but things like \vskip would be out. We might also need to consider modifications of the LaTeX syntax to aid the conversion to XML based formats, e.g. this is done in gellmu, but in such a way that it is easy for LaTeX aficionados to adapt.

Answer 2

In the context of this Pete Sefton's response to my original post is very interesting as well describing a Java plugin to really make the web properly editable. Slightly different direction again but worth considering.