[XQuery] Path expressions and axis features

Reading the working draft (W3C Working Draft 12 November 2003) and
testing some XQuery tools left the impression that the axis features are
sort of neglected by leaving features optional, which is not acceptable.

The introduction of the Full Axis Feature (section 2.6.3) open a way for
implementers to claim XQuery conformity without supporting certain axis
such as the sibling and ancestor axes.
These might not be relevant in unordered databases such as address
tables, where sorting takes place, if required, after querying; but XML
data provides an essential component namely the preservation of the
order. 

Annotation formats for speech for example rely on rather flat XML trees
that maintain the linear order of speech. These annotations clearly show
a data centered approach to XML, which makes it a candidate for
processing with XQuery rather then XSLT. But querying such data is
interesting if access not to the whole subtree but to a context is
possible.

The same is true for flat structured textual documents where
distributional analysis of texts are required and wanted. 

Accessing the linear structure of XML documents is a major advantage of
XML based databases for certain applications and clearly distinguishes
these DBs from relational data where these access structures are not to
be maintained. 

Especially for the processing of natural language, in data or document
based XML, the full set of axis needs to be available. 

I would strongly recommend the change of the following:
1. Section 3.2.1  

OLD sentence: XQuery supports the following axes (subject to limitations
as described in 2.6.3 Full Axis Feature): <bulletlist>

NEW sentence: XQuery supports the following axes: <bulletlist>

2. Section 2.6.3 needs to be deleted. Vendors will provide their
personal flavors of XQuery, if the feature is in the specification or
not. But at least they should not find an excuse for omitting certain
axis that are justified. 

If necessary I can provide example queries which are needed in language
databases.

A workaround with user defined functions (using features such as the
node number or position) or more complex contstructs is not acceptable,
as it seems to violate the priciples of navigation in XML trees.

Missing features like this would doom XQuery to failure in large and
progressing XML using communities. 

Thorsten

-- 
-----------------------------------------------------------------------------
///////// Thorsten Trippel         ttrippel@spectrum.uni-bielefeld.de
   //     Computational Linguistics and Spoken Language
  //  //  Research Group: Text technological modelling of information
 //  //   URL: http://www.spectrum.uni-bielefeld.de/~ttrippel
    //    Bielefeld University, Federal Republic of Germany
///////// Department of Linguistics and Literary Studies     
-----------------------------------------------------------------------------

Received on Thursday, 15 January 2004 05:20:16 UTC