In message <0DB95D5E-F2EB-4574-A288-F5D45C56D5E2@indiana.edu> "John A. Walsh"
<jawalsh@indiana.edu> writes:

> As a Java app, it would be easier to embed a Java XQuery solution  
> (rather than something like bdbxml).  I wonder if just dropping in  
> the saxon8 jar and reading the ODDs off the file system would give  
> Amit the XQuery functionality and performance he needs for Byzantium.

It would all depend on whether there was any cacheing happening.
If Byzantium had to read the whole TEI P5 each time it wanted to
ask ¨which attributes does this element have ¨, then it could get
painful. In the XSLT underneath, the first document() appears to
cause a cache, so subsequent hits in the same XSL are fast. 

for some questions, you can read the individual file containing
the elementSpec, of course. But for some other questions, you cant.

What one would want to avoid, of course, is the Byzantium code maintaining
its own data structures.
-- 
Sebastian Rahtz    
Oxford University Computing Services
& JISC Open Source Advisory Service