[Refdb-users] xml and db storage

[Bruce D'A. <bd...@fa...>]

I've got some ideas I've been pondering, that are not necessarily 
specific to RefDB, but obviously related:

As you guys know, I've been playing with a Java-based XML database.  
I'm pretty well convinced of the advantages of being able to dump xml 
whole into a db and run random xpath queries on it.  I know, however, 
there are advantages to using traditional RDBMSs, particularly if you 
need to scale to using many users.  So, I wonder what you -- 
particularly Markus -- think of the following on adding this 
functionality to postgres?

http://www.throwingbeans.org/tech/postgresql_and_xml.html

How do I envision applications?  Well, it's an obvious way to get easy 
support for MODS into RefDB.  No need to carefully map all of the xml 
structures to tables and rows, and then back; just dump the xml in and 
suck it out when needed.

Also, I'm big on annotations, as you also know.  Whatever happens with 
our DocBook proposal, I really want to be able to have notes marked up 
in ways like this, and to be able to run queries on that marked up text:

   <note user="darcusb">
     <para>Here we have a paragraph with a:
     <blockquote reference="Doe2002c" related="Jones2000b" 
keywords="philosophy culture" start="21" stop="22">
       <para>As with the quote element, blockquote can include attached
       citation information. In this case we have the single reference
       with a page span, along with a "related" reference that links
       this quote conceptually with another reference.</para>
       <para>We also add here an <emphasis source="original">emphasis
       tag</emphasis>. The attribute here indicates whether the
       emphasis has been added by the note-taker, or was present in the
       original source.</para>
     </blockquote>
   </para>
</note>

Or this:

     <notes>
       <note>
	<para>Article deals with consequences of domestic post-9/11
	war on terror.</para>
	<para><quote keywords="anti-terrorism detention" source-name="DoeJ">By 
my count, they've detained over 5,000 foreign
	nationals in anti-terrorism-related initiatives</quote></para>
       </note>
     </notes>

In this case, the quote is from someone referenced in a newspaper 
article.  I want to note who it is they interviewed, and add a note to 
that person's record.

So imagine here a GUI-ified representation of this -- a web client:

I type "anti-terrorism" and "detention" into a keyword field.  The 
latter entry comes up, and I am presented the quote that is hot-linked 
to the name record (an xslt could render the relationships clearly).  I 
click on that, and find the person, their description, and perhaps 
contact info, as well as other records they may be referenced in.

With MODS, I could imagine this all done in XML.  You have MODS files 
for the records proper (with a .mods extension), for names (.names), 
for maybe serials (e.g. journals), and then separate files for notes.  
All of these are then cross-linked, and I can add names or notes 
independently of bibliographic records per se.

What I am wondering if with RDBMS + XML DB functionality this would be 
a good idea too?

Bruce