scheme banishes browser plug-ins
Ted Smalley Bowen,
Technology Research News
There’s not much guesswork involved in
pulling a book off the shelf. Little has changed since Johann Gutenberg
came up with movable type -- provided you read the language, you can just
crack the cover and you're on your way.
By contrast, the babble of data formats represented on the present-day
confuses the process of accessing digital
information. Your basic Web
browser can only handle so many data types, and the prospect of searching
for and adding the right plug-in can be laborious even when successful.
And as with all things digital, nothing stays the same for long.
In order to display what you want, your browser must be able to make sense
of the relationships within groupings of digital files so it can, for
instance, show the correct graphic with a block of text, accommodate both
the thumbnails and larger views of a set of pictures, or synchronize a
video with lecture slides. The problem involves finding and coordinating
the right programs to display the various types of text, image, or sound
files scattered throughout the Internet.
A Cornell University researcher has found a way to identify key characteristics
of digital content in order to match content with programs that can display
it in a browser.
The scheme involves a modification of existing software for storing and
displaying digital content files that separates the two processes, according
to the researcher, Naomi Dushay.
The scheme is especially useful for displaying content from the Internet
because neither the content nor the program that activates it need to
be present on the system that displays the content.
The context broker software at the heart of the scheme is a set of Java
programs that generates Web pages. The context broker acts as a go-between
for the repositories that store digital content, the programs that act
on the content, and the browsers that display the results.
By separating the storage and maintenance of digital content from its
presentation, the scheme could foster more specialization throughout the
digital community, said Dushay. “Digital content providers might choose
to specialize in content only, or also put up context brokers and go after
both presentation as well as content,” she said.
The scheme also opens up the possibility of more individualized presentation
of data, or for augmenting the presentation of data created by others,
said Dushay. "For example, the Cornell University Library might have some
whizzy presentation, rendering mechanisms targeted for members of the
Cornell University community. These might be made available via... context
The method could also allow "searching, categorizing sites such as Google
or Yahoo [to] provide a context broker so users want to access resources
via their sites," she said.
The context broker gains information about the content from the metadata
contained within content files. Metadata is data about data, and can include,
for example, descriptions of the contents of a file or groups of files,
or administrative details related to the data.
The scheme uses this structural metadata to identify the appropriate program
for presenting the data. The programs are listed in a behavior registry,
which also includes information about how a playback program can be accessed,
which data structures the program can handle, and what effects each program
The context broker ties digital content to the behaviors these playback
programs can produce. By changing the behaviors listed in the behavior
registry, content behaviors can be changed without modifying the content
Dushay tested the scheme using the Cornell Digital Library research group’s
Fedora, a repository that stores agglomerations of different types of
data drawn from multiple locations.
To use the scheme, a user looks through a list of playback effects available
for a given piece of content in the repository and requests that a certain
program present the content in question.
The software matches the content's structural metadata and access points
for assigning behaviors to the content with the appropriate playback program,
then loads the program and uses it to access the content and display it
in a Web browser.
A lot of content is now created with the kind of explicit structural metadata
the scheme calls for, said Dushay. In addition, objects lacking it could
be assigned metadata by inferring the information or by "using some sort
of fuzzy pattern matching on structural access points required by behavior
mechanisms,” she said.
Although the context broker model does not require control over the content
or playback programs, end-users will need direct or indirect authorization
to access the content. Metadata access could be made separate from access
to the data itself, said Dushay. “It's possible to expose structural metadata
without exposing the content itself. It's possible to determine the potential
for [playback] behaviors with only the structural metadata, though eventually
that content will be required to actually perform those behaviors,” she
Dushay is also planning on making the scheme work with other content repositories.
The scheme will eventually use more sophisticated pattern matching as
a means of sorting through structural metadata, and there may be ways
to add more detailed descriptive information to that metadata, she said.
She also has plans to tailor the context broker’s playback for individual
users, to allow differences in spoken language or language proficiency
to condition how each user receives the data.
To bring the scheme beyond the proof-of-concept stage the amount of computing
and network resources needed to pull the various pieces together could
become an issue, Dushay noted.
The network resources needed to provide access to structural metadata
"could get costly, but perhaps this could be alleviated with caching or
mirroring of frequently used data and mechanisms at context broker sites,”
The format of metadata and the behavior mechanism input requirements will
also impact performance, she said. "If it's possible to index the input
requirements [and] structural metadata for fast look up, great. But if
they're "fuzzy" matches, then this may become a performance issue.”
Dushay's work was funded by the National Science Foundation (NSF).
TRN Categories: Databases and Information Retrieval; Internet
Story Type: News
Related Elements: Technical paper, “Using Structural Metadata
to Localize Experience of Digital Content”, arxiv.org/ftp/cs/papers/0112/0112017.pdf
Shake and serve
Odds not hopeless
for new Web sites
banishes browser plug-ins
Polarized light speeds
File compressor ID's authors
Research News Roundup
Research Watch blog
View from the High Ground Q&A
How It Works
News | Blog
Buy an ad link