Hi Patrick,
On Aug 10, 2006, at 4:58 PM, Patrick Durusau wrote:
> One of my library contacts at the Extreme conference has suggested
> that if we are going to standardize on bibliographic data that we
> should use MODS (Library of Congress).
Yeah, of course a library person would say that. Talk to an archivist
and you'll get a different answer.
> If that would work it would save us from having to write such a system.
I could write a book on what I know about MODS and related stuff, so
will just try to boil it down to no, NO, NO!! You (and many of the
librarians) are not thinking about this problem right.
I say this after having spent massive amounts of time working with the
Library of Congress on improving MODS, and subsequently promoting its
use. Really, I spent far more energy promoting MODS outside the library
world than the librarians have. I even at once wanted to include it in
ODF.
I was wrong, and I'll try to just give you a rough outline of an
explanation ...
> Are there any downsides to using it?
Yes.
1) the schema is needlessly complicated in many places, and too limited
in others
2) it is inconsistent, which has other side-effects (see below)
3) it is all based on XML schema; a completely monolithic solution
4) because of 3, awkward to extend
5) also because of 3, really loose
6) not designed for this use case (missing some pieces; see above)
7) because of 2, linking is awkward and limited
8) again, why use one framework for a document and another for the
bibliographic sources, which themselves are documents?
I could really go on and on about the problems with it for this use
case. And if you read my blog, you'll note there's probably nobody in
the world that knows more about MODS than I outside of the Library of
Congress itself.
Last year I stood up before a room full of library people at a
conference (where I was invited to speak) and told them their XML
schemas (I meant MODS among them) were over-the-top bloated, and they
needed to get serious about RDF. I showed them an example.
Why on earth does it make any sense for a community (the librarians) to
invent monstrous schemas like this when they could have and should have
done it in RDF? DC and DCQ covers so much of what's in MODS, but
because it didn't cover some other things they did need, they went off
and invented their own schemas.
And when I and others from the citation focus say "but we need x, y, z"
the Library of Congress has to approve all of that, and they really
don't care about some of our needs.
Monolithic schemas of this sort are the information equivalents of
dinosaurs.
This is EXACTLY the problem we need to avoid. I'm absolutely serious;
this approach is broken.
My own summaries of the issue related to the conference talk:
<http://netapps.muohio.edu/blogs/darcusb/darcusb/archives/2005/10/18/a-
complete-metadata-cycle>
<http://netapps.muohio.edu/blogs/darcusb/darcusb/archives/2005/10/22/
xml-and-rdf>
... and a few (brief) of responses (to the talk):
<http://curtis.med.yale.edu/dchud/log/conferences/access-2005-
darcustalk>
<http://blog.uwinnipeg.ca/loomware/archives/001466.html>
<http://www.wallandbinkley.com/quaedam/?p=43>
Note: WRT to Rob's question today, I only come to RDF out of practical
frustration with the alternatives. My book was authored using MODS as
the bibliographic data source, and when I finished and went through
trying to clean it up (in part by normalizing), I realized it was
hopeless. I've converted it all to RDF and haven't looked back. And I
don't even typically use RDF tools!
> I am thinking that it might get us some mileage in the library
> community.
The LoC is actively working with MS on the ECMA spec. I invited them to
join the ODF metadata SC. I don't draw a very positive conclusion from
that.
> On the other hand, we should mention that the NLM (National Library of
> Medicine) metadata system may be preferred by some users.
Exactly; there are probably 20 bibliographic schemas out there. The
ONLY sensible thing that will actually work is to do as I've been
suggesting. Forget monolithic schemas developed by particular
communities, and instead think "modules."
Am working on a demo of sorts using XSLT ;-)
Bruce
