Metadata for a Better Information World
Leslie L. Daigle
May 8, 1996.
Bunyip Information Systems, Inc
Internet Information Transactions
Basically based on:
- a client's information need
- an information processing task (activity)
- available information resources
Typically, processing is shared between the client and server
ends of the transaction:
- server manages resources
- client selects and renders resources
(N.B.: very little sophistication in terms of the information processing
task)
Current Trends
An increasing amount of the transaction processing is being held at
the server end:
- Yahoo, specialized search servers
- Java, active web pages are rendered at the client end, but essentially
represent a view held at the server's end of the transaction
This yields an imbalance in the information transaction model.
Hopes for the Future
Clients need to be able to reclaim some of the processing task supporting
the information interaction.
- client must be able to represent its information need to information
servers
- services must be able to provide descriptions of their resources in
some standardized form
The content of these resource descriptions
- will vary significantly from application to application (e.g., music vs.
geospatial data)
- should share a common carrier structure/format
A Perspective on URCs
URCs are:
- a representation of an object or service
- first-class information objects in their own right
Therefore:
- are not inherently active
- are independent of any particular information service
What People want from Internet Resource Description
URCs are many things to many people. Roughly, their perceived uses can be
categorized as follows:
- carrying specific metainformation to distinguish between resource
instances (for URN resolution -- the original reason for which URCs were
conceived)
- list of published or mirrored instances
- location of instance
- format of instance
- cost of instance
- more service-specific information
- information to support cacheing systems
- history of document usage
- general cataloguing information
- application- or service-specific metainformation
- e.g., melodic contour representation for a specific
music-retrieval system
- semantic weightings for a particular text-retrieval system
- labels for a ratings service
URCs and Metainformation
URC design to date:
- has focused on trying to construct a unified structure (with canonical
representation) for all applications
- attempts to be flexible by permitting transmission in any one of several
formats.
Originally intended:
- for URN resolution, or service-specific resource characteristics
- where the boundaries of the problem space are known
Now suggested for use:
- for application-specific metainformation
- where the boundaries of the problem cannot be known in advance, because
the potential applications are limitless
Can URCs Deliver?
The unbounded problem of generalized application-specific resource description
- has lead to a weighty design of URCs
- makes them awfully complex to use for the simpler, bounded problems
- must be solved
What needs to happen:
- recognize the difference in nature of the problems
- solve the resource representation problem for Internet (bounded) problems
- provide gateways to more application-specific represenations
What this means for Information Transactions
- Clients can use URCs to assess/describe Internet resources from a
mechanical level (i.e., more meaningful than filenames, etc)
- The URC framework can serve as a gateway to more application-specific
representation schemes, usable by some clients and not others