XML + Elementree + (ZODB or Durus) experiments?

[News Reader]

A year or so ago, there was a posting - I believe on someone's blog - 
which told of a unique experiment.  The experimenter tried loading and 
searching a large XML document based on three strategies:

1. (I think) elementree directly
2. Store entire XML document in ZODB (or Durus, can't remember)
3. Subclass Elementree.Element, store each XML node separately in the 
ZODB/Durus.

The experimenter then did timings on opening up and using the XML 
documents in each of these three ways.

Does anybody remember that post?  Does any body have a link, code, 
cache, anything?

Please reply to the list, as the email address is invalid.

Thanks,

E. List