Many ILSes give the ability to export item data in the MARC record in a 9xx tag, (usually the 949 since B&T and other book jobbers like to put holdings data for newly-acquired items there so the ILS can automatically create an item record when the MARC record is loaded).  That is how I've been getting location/collection code info into my Solr-based catalog.  So you might want to look into that.
 
I think having separate XML files for holdings data (and hence, a second install of Solr just for holdings data) is less than optimal for a myriad of reasons.  Likewise I think XSLT is a pretty poor tool for generating Solr records.  XSLT is really difficult to do the kind of data manipulation I've been finding I need to do on our MARC records to get them nice and Solrized.  Also, very very poor performance.
 
--Casey

>>> [log in to unmask] 1/17/2007 12:48 PM >>>
On Jan 17, 2007, at 2:26 PM, Andrew Nagy wrote:

> Nate, it's pretty easy.  Once you dump your records into a giant marc
> file, you can run marc2xml
> (http://search.cpan.org/~kados/MARC-XML-0.82/bin/marc2xml).  Then
> run an
> XSLT against the marcxml file to create your SOLR xml docs.

Unless I'm totally, hugely mistaken, MARC doesn't say anything about
holdings data, right? If I want to facet on that, would it make more
sense to add holdings data to the MARC XML data, or keep separate xml
files for holdings that reference the item data?

In a lot of cases, location data might not be a hugely important
facet; at Madison, we have something like 42 libraries spread thinly
across campus (gah!) -- each with different loan policies -- as well
as a few request-only storage facilities. So there's a lot of "Stuff
I Can't Check Out" and a lot of "Stuff I'll Need To Wait For" in our
collection.

Thanks!
-Nate