On Tue, 2010-10-26 at 03:32 +0200, Alexander Johannesen wrote:
>    "Here's our new thing. And we did it by simply converting all our
> MARC into MARCXML that runs on a cron job every midnight, and a bit of
> horrendous XSLT that's impossible to maintain."

I am in the development department of our library. We're a diverse bunch
of guys, ranging from the bottom (that's me, hacking Lucene) to the top
(our graphics guy). Somewhere in the middle we have 2 librarians. They
do not program in traditional languages, but have been trained to
produce XSLT's and it actually works! They are capable of translating
their vast knowledge of the myriad of standards we encounter into code
that transforms our XML-input into something we can use for indexing.

"Aha!", you counter, "why not train them to use X instead, since X is
much better at transforming normal MARC?". The answer is that MARC isn't
the only format they need to handle. We currently have 20+ different
sources that they need to transform. All of them except one is XML. The
one is ISO 2709 MARC, which we - naturally - transform into MARCXML so
that it can be processed the same way as the rest. There might be better
tools than XSLT for transformation of XML that we could use, but the
XML-part is so ubiquitous at this point in time that it is the obvious
choice for common ground.

MARC is just one in many. It might be the most evil and unruly beast of
the bunch, but we tame it with the same tools as the rest.