texi to epub

[Kurt Hornik]

Friends,

You may be aware the R (<https://www.R-project.org>) has been using
texinfo for its basic manuals "forever".  A long time ago, it was
suggested to us that ebooks were convenient to read, so we looked into
coming up with a pipeline to turn .texi to .epub.  See
<https://cran.r-project.org/manuals.html> for the current results.

This pipeline is very simple: we first convert .texi to .html, and then
use Calibre's ebook-convert to convert .html to .epub.

This has worked rather well for us for a long time.  A few months ago,
we got a bug report that this would give invalid epub according to
epubcheck, see <https://bugs.r-project.org/show_bug.cgi?id=18182>.

My current status is:

* The texinfo TODO has

    Support the epub output format, http://en.wikipedia.org/wiki/OEBPS

  and there are 

    <https://lists.gnu.org/archive/html/bug-texinfo/2018-07/msg00009.html>
    <http://per.bothner.com/blog/2016/texinfo-roadmap/>

* Calibre ebook-convert will not make attempts to create valid epub
  according to epubcheck.

* Dr Google finds

    <https://github.com/ikrukov/epub>
    <https://github.com/jlhg/texinfo2epub>

  which suggests using a pipeline using makeinfo --docbook and then
  xsltproc.  There is also

    <https://opensource.com/article/17/9/docbook>

  which suggests using a pipeline using makeinfo --docbook and then
  pandoc --from docbook --to epub3.

  I can get both approaches to "work" in the sense that epubcheck is
  happy, but the visual quality of the epubs is not so good as the ones
  I get using the current pipeline.

  Which may be a bit of a self-fulfilling prophecy, as I use Calibre for
  reading the epubs, and have no idea what others are seeing with other
  epub readers.

So after playing with this a bit, I am now wondering how to best move
forward.  In an ideal world, the Texinfo TODO

    Support the epub output format, http://en.wikipedia.org/wiki/OEBPS

would be DONE :-)

In the discussions in the bug report for R, someone wrote

  calibre does not guarantee that an EPUB produced by it is valid. The
  only guarantee it makes is that if you feed it valid XHTML 1.1 + CSS
  2.1 it will output a valid EPUB.

and of course makeinfo gives HTML 4.01 Transitional: I also tried the
effect of going through HTML tidy to turn that into XHTML, but that did
not make epubcheck happy.

Any ideas/suggestions/... how to best move forward?

Best
-k