From: Aaron Chang (anc327@yahoo.com)
Date: Mon Apr 19 2004 - 03:00:08 PDT
Review 2
CSE 544
1. central conclusion of paper: the traditional
approach of using RDBMS systems plus new algorithms
can be used effectively for some, not all, XML
document processing/querying; the authors appear to
show a balanced study of the pros and cons using the
conventional methods
2. part of the difficulty: xml is semi-structured,
variance in complexity of DTDs
3. method for data flow: DTD->relational schema->load
tuples->translate xml queries to sql queries
->convert retrieved data back to xml
- seems pretty straightforward. not too surprised
this works for the most part
4. methods for storing xml in relational format:
- tranforming/simplifying DTDs
- basic inlining: use of DFS graph processing is
pretty interesting, but point out rightly
that this kind of processing is good for only
certain types of queries
5. shared inlining technique:
- looks for shared element nodes in multiple tables
- key idea: element nodes partitioning into tables
determines schema
6. basic, shared, and hybrid inlining were tested by
measuring # of joins required for a given query
7. results:
- # of joins scale with path length
- performance gap between shared and hybrid depend on
path length
- not too surprising, gap difference is rather small
too
8. query conversion from path expressions to sql look
pretty straightforward, seems like
with xml, root path checking is critical
9. converting sql results back to xml is mostly the
same as it's currently done it appears
__________________________________
Do you Yahoo!?
Yahoo! Photos: High-quality 4x6 digital prints for 25¢
http://photos.yahoo.com/ph/print_splash
This archive was generated by hypermail 2.1.6 : Mon Apr 19 2004 - 03:00:16 PDT