Transcript PDIS Steering Group

Mobile Data Management
Ken Rimey
2003-2004:
• Personal Distributed Information Store (PDIS)
2005-:
• Too Many Boxes - end-user digital asset management for
networked consumer electronic devices.
• µXMLDB - an XML database for mobile devices.
PDIS (2003-2004)
• Next-generation data synchronization: Enable people
•
•
to store copies of their data on several devices, and
several computers, and to keep these copies in sync.
State-transfer, update-anywhere replicated XML
database.
– PIM data.
– Digital media metadata.
Ported to Nokia’s Python for Series 60.
TMB: Too Many Boxes
• Ken Rimey, HIIT
• How will consumers manage their files on multiple
•
•
•
handheld devices, computers, and home/car
entertainment boxes?
How to combine control with simplicity, rejecting both
complete automation and complex graphical user
interfaces?
Utilize PDIS/PDOS results to build a home network
testbed.
2005 volume: 4-5 person-years
Hypothesized Issues
• WHERE to store copies of files?
• WHEN to synchronize them over wireless networks?
• WHOse personal devices should have access?
TMB Methodology
• Build and study three prototypes:
– Start with raw sharing of files among several phones
and several computers, presumed to belong to the
same person (1H2005).
– Then add in multiple people (family members, etc.),
end-user configurability, a focus on specific media
types, a mockup of a home entertainment system,
and some experimentation with gestural or
multimodal interaction techniques (-2006).
– Finally, capitalize on this infrastructure by integrating
additional applications or devices, or by adapting it to
some other setting besides the home (2007).
µXMLDB: An XML Database for
Mobile Devices
• Kenneth Oksanen, HUT and HIIT
• XML is of increasing interest as a vendor-neutral
•
•
•
exchange and storage format in handheld devices.
Existing XML database solutions target large,
administered servers and mostly text-oriented XML
data.
Needed: A self-administering XML database with
small memory and flash consumption, a simple
vendor-neutral index creation API, and extensions for
querying data-oriented XML.
2005 volume: 3 person-years
Research Topics
• Index management
• Query language
• Representation of stored XML
Index Management
• Indexing of XML is not well understood.
•
•
•
– Some systems build large generic indices.
– Others allow specifying more accurate indices, but
no standard API exists for this.
Specify intuitively with query templates:
/todo[date="%s"]
Or automatically based on queries:
/todo[date="20040922"]
/todo[date="20040923"]
The latter leads to an opportunity to construct very
concise and selective indices.
Query Language
• Based on XPath and possibly XQuery.
• Extended for, among other things, categories:
•
– The query /todo[isa(@category, "work")] matches
the document <todo category="PDIS">...</todo>,
if “PDIS” is defined as being work.
– Categories change, and we don't want to reindex
the whole database when they do.
Handling RDF and OWL is a possible further research
issue.
Representation of Stored XML
• Allow use of compression.
• Store big CDATA blobs separately.
µXMLDB Deliverables
• An open-source C implementation for Unix and
Symbian OS.
• SWIG interface supporting other languages.
• Use existing system software (Berkeley DB, SQL, …)
for persistence, storage and indexing.