Transcript Asilomar What about Xquery?

XML
(with a bias towards query language issues)
A boring research topic?
A new frontier?
A means to keep standards people busy?
Prepared by S. Abiteboul and J. Widom
XML
• Rapidly adopted by industry
• A format for exchange of small/medium pieces of data?
– Yes
– But when archived, grows to large volumes
• Also a data model?
– Yes, for all kinds of data
– From unstructured documents to collections of structured data
– Warning: this is not the relational world!
• Permissive typing, full-text search, …
The database community should be very involved,
perhaps very concerned
Some XML Issues
• Storage of XML
– Native vs. XML-relational
• Lesson from OODB: it’s not only a technical issue but a business one
• Situation is different: more $ involved
– Efficient representation, compression
– Key issue: interface
• DOM, SAX, query language, DB-like API, …
• Revisiting old topics
–
–
–
–
–
Database design
Integrity constraints
Concurrency control
Access control
Etc.
All topics are under active investigation,
sometimes reinventing the wheel
Universal Query Language for XML
•
Problems with XQuery
–
–
–
•
Undermining XQuery with something better?
–
–
•
Focus on complex queries; really need simple filters + IR-style search
Too complex, too ambitious, lack of underlying paradigm, too much
politics!
Too broad spectrum of applications? From documents to data
Personal viewpoint: Small core OQL-like + functional plug-ins
Too late? We need a standard now!
What about updates, standing queries, deltas,
constraints, …?
This direction has been mostly deactivated by XQuery
Scientific:
Is Xquery good or bad from a scientific viewpoint?
Politics:
Should we push for XQuery?
Back to Basics: Query Optimization
• For subsets of the language
• The tree structure is definitely a new ball game
– New index structures
– New cost models
– New everything
• Depends on storage
– Relational, native, others
• Revisit old problems
– Distributed query processing
– View maintenance
All topics are under active investigation
(but more effort on distribution wouldn’t hurt)
Back to Basics: Foundations
• Lots of work on semistructured data
– First-order logic and relational languages: strong
– OQL/functional languages: reasonable
– Full-text search: messy
• Significant issue: typing
– Much more complex than in relational world
– Not settled (XML Schema, tree automata, …)
– Query type-checking, type inferencing, update consistency
• Very active area
– People from database theory, functional programming, automata
theory, …
All topics already active, not simple, require more work
The Real Frontier
(The World is Changing)
OLD data management
NEW data management
•
•
•
•
•
•
•
•
•
•
•
Closed world
Client/server
Distributed databases
Query/answer
• Active databases
• QBE interfaces
Openness
P2P applications
Web-scale data
Subscription queries
Queries over streams
ADB + Web services
New interfaces
Research should focus on the new issues rather than
on traditional processing of single-site XML data
Beyond XML: semantic Web
Discussion