The reason for my question was that there are still some cases where pandoc-citeproc does not sticks to csl-rules (e.g. https://github.com/jgm/pandoc-citeproc/issues/79). And I thought it might be easier to use citeproc-js, which should be fully csl-compliant. But, if I understand your answer correctly, this will only lead to other problems.

Am Samstag, 28. Oktober 2017 00:22:50 UTC+2 schrieb John MacFarlane:

Several reasons:

1.  citeproc-js would be an external dependency that users would
have to install.  pandoc-citeproc is a Haskell library we
can just compile in.

2.  pandoc would have to shell out to call citeproc-js
(which, earlier, didn't even have a way to call it as a
standalone program)

3.  pandoc-citeproc works with native pandoc types.
With citeproc-js, we'd have to convert to citeproc JSON,
and then convert back.

4.  pandoc-citeproc handles math better (because it was
designed to work with pandoc's math support).


+++ Denis Maier [Oct 27 17 06:55 ]:
>   I was wondering why pandoc does not use citeproc-js but its own
>   pandoc-citeproc. As far I understand citeproc-js is the canonical
>   csl-implementation, and, as such, fully compliant with the
>   csl-specification.