Hi, sorry for the delay, I was away for a while.

> I'd suggest 2 :) But not that I don't use this feature.
>
> It should be easy to unify the code: something along the lines of starting the process, and then looping over bibtex files and sending them one by one to bibtex2html's standard input.
>
> Cheers,
> Clément.

Please find attached a tentative patch using `process-send-string'.  It seems to
work on my test cases, but I don't know how to wait for the bibtex2html process
to finish before processing the output.  I am currently using a while loop (see
l. 96 of the patch or l. 256 of the patched ox-bibtex.el) combined with a
sentinel changing the while condition upon completion.  This seems suboptimal
but I don't know how to achieve that more elegantly.

I still have a few questions:
1. How can I wait on the subprocess to complete after all the bib files have
   been passed via stdin?
2. Why is this approach preferred over concatenating bib files and delegating
   processing to the bibtex2html executable?

Thanks for the help,
thibault