Aaron Ecay <aarone...@gmail.com> writes:

> Hi Richard, hi all,
> 2015ko urriak 26an, Richard Lawrence-ek idatzi zuen:
> [...]
>>> I was working on this rather intensively at one time, but I had to stop
>>> because other aspects of life intruded.  I have just been coming back
>>> towards a situation where I can imagine myself having some (still small,
>>> but non-zero) chunks of time to devote to working on org.  So I hope I
>>> will be able to pick this back up, but (regrettably) I’m not able to
>>> make any promises.
>>> Based on my recollection, here’s what the problems were when I stopped:
>>> - The only “off the shelf”-capable citation processing library that we
>>> found last time is in Haskell, which introduced some difficulties for
>>> distributing the resulting tool.  I know some projects
>>> (e.g. git-annex) are written in Haskell and distributed as static
>>> binaries for windows/mac/linux/etc.  We’d need to figure out how to do
>>> this, or find another citation processing library in an
>>> easier-to-distribute language.
>> Yes, this is my understanding, too.  In particular, there does not seem
>> to be an Elisp CSL library, and it would be a lot of work to write one.
>> The other CSL library that looks complete and usable is citeproc-js; but
>> like the Haskell library (pandoc-citeproc) it would need to be wrapped
>> somehow so that it can talk with Org.
>> It should be relatively straightforward for someone who knows Javascript
>> to write such a wrapper, if anyone wants to work on that.  But this does
>> not really solve the problem with distribution.  
> It solves many of the hard problems though.  Node.js is distributed
> as a binary for many platforms.  We’d just have to direct users to
> install this in the “normal way,” and use the installed binary to
> interpret the JS source.  Whereas for haskell we’d be stuck building
> the binary ourselves, worrying about static linking/dll hell/32-bit
> dinosaurs/any of a half-dozen other problems that I don’t really
> understand.

I would feel more comfortable relying on a JS library.  Perhaps it’s also
easier to find people who are willing to work on/knows JS over the long

> OTOH, pandoc-citeproc includes a bibtex parser; we’d need to write a JS
> one and wire it up to citeproc-js.  When I looked (quite some time ago),
> there did not seem to be any good bibtex parsing libraries in JS (and
> several third-rate ones).

Bibtex support is essential, of course.

Can someone remind me why citeproc-java isn’t good?  AFAIR, it has a
bibtex parser.  But probably it lacks in some other dimension...

> OT3rdH, responding to Matt’s message
> <http://mid.gmane.org/can_dec_52sp6ghr56pudhih69ksprq0vdw2zmnp5799-cta...@mail.gmail.com>,
>> The disadvantage is that, from what I can tell, the javascript
>> implementation is the canonical version of citeproc, and the place where
>> improvements are pushed first.  So, for instance, if one wanted to
>> implement an org-syntax output format for citeproc, citeproc-js would be
>> the most likely project to support that work.
> Pandoc can output org syntax, so it may be that we can just link with
> the main pandoc haskell library as well as pandoc-citeproc and solve
> this ourselves, without needing upstream support.

Do we WANT to depend on Pandoc?  I would say "no".  In my OS, where we
finally got a binary distribution of pandoc, the size of pandoc is still
1600Mb!  I don’t know if this is representative of other systems, though.
E.g. what is the size of pandoc+deps in Debian?


Lasciate ogni speranza o voi che entrate: siete nella mani di'machellaio

Reply via email to