Updated experimental JATS-based research document system

[castedo]

I have deployed a new version of my experimental research document system, based on JATS XML and pandoc markdown. Here is an example of a document: https://castedo.com/doc/151

Some features that might be interesting:

• the web page view is optimized for encouraging communication (email, gitlab or hypothes.is)
• both the PDF and web page are generated from the posted JATS XML file
• the JATS XML file is separately generated from author source code, mostly pandoc markdown

Some limitations I have hit that might be of interest:

1) Hypothes.is did not handle updates to the document very well.

An annotation that I added became a nonsense annotation two versions of the document later. Details are here: hypothesis/client#258 This is a natural limitation of public annotations. Without assistance from humans, it is very hard for a computer algorithm to automatically know when an annotation is no longer a match for new text on a page. My plan is to switch to using a Hypothesis publisher group rather than the public group, so that annotations can be managed for a more sensitble reader experience.

2) There is a limit to how much JATS specific handling pandoc can parse out of arbitrary JATS XML.

I quickly hit this limit attempting to do structured abstracts. I can go into details if anybody is interested. I conjecture that pandoc will still be great for parsing text inside JATS XML, like the body with math inside. But I speculate a library like pypi.org/project/elifetools/ will always be ahead of pandoc in handling general purpose JATS XML. My plan is to try using github.com/elifesciences/elife-tools but perhaps use pandoc for handling content inside like the body and abstract.

If anybody has suggestions or feedback, please do share.

For my next upgrade I plan to deposit the JATS XML with Zenodo while keeping that content and the presentation forms of HTML and PDF separate.

Lastly, if anybody wants to potentially use the source code, I plan to open-source the useful parts. I haven't yet because it's embedded in some very experimental code that generates my website.