Ulf Martin
I think the crucial point for any TeX community is the ability to use the rather huge amount of BibTeX legacy DBs.
How about the state of CSL (or RDF) to BibTeX converters?
I don't care about BibTeX myself, so such things aren't my focus. However, I think a good XML/RDF data format makes it pretty easy to downconvert to formats like BibTeX. Indeed, it took me 30 minutes or so to write a decent XSLT to convert MODS to the RDF/XML I'm using. That was only targeted at book descriptions, so it would take more time for a comprehensive version, but it shows it's not hard. The hard part, in fact, is the logic for conversion, and most of that is clearly documented in the bibutils source code.
bibutils uses MODS as its native intermediate format and converts from and to BibTeX (not always 100% correct, though).
Correct, though it's actually more complicated than that. It uses a C-based internal format that is based on lessons from MODS and from converting the other legacy formats.
Summary -------
So, at present we already have:
(1) MODS <-(bibutils)-> BibTeX -(bibmod)-> ConTeXt
For an XML-based format in a ConTeXt context we would like to have:
(2) BibTeX <-(a)-> XML -(b)-> ConTeXt
*We* wouldn't include me. I deal much more with RIS or Endnote formats than I do with BibTeX. But I don't use ConTeXt for authoring either ;-)
using the rather nice XML processing capabilities of ConTeXt for step (b).
Now, there is an XML markup for BibTeX: BibTeXML http://bibtexml.sourceforge.net/ This isn't too bad, in my experience (it is, at least, lossless, contrary to bibutils). Thus
(3) BibTeX <-(bibtexml)-> BibTeXML -(b')-> ConTeXt
would be an instance of (2).
Yes, but BibTeXML still has all the problems of the BibTeX model.
CSL could use XSL transformer:
(4) BibTeXML <-(XSLT)-> CSL -(b")-> ConTeXt
All CSL is is a language-angostic XML config language. You could write a CSL engine in whatever language you want: TeX, Lua, Perl, Ruby, C. *I* wrote mine in XSLT 2.0, but that's mostly because of limited skills with other langauges. I also designed citeproc, BTW, to have both an input and output driver system. So while I use an RDF/XML representation internally, it wouldn't be too hard to write other inout drivers. A next-generation mbib module probably ought to do the same, so that while it might have a richer core format, it could still be fed BibTeX, or even MODS. Bruce