Metadata
for Client-side Content Adaptation
Oskari Koskimies
Nokia Research
Center
Disclaimer: Opinions and views expressed in this paper belong to the author and do not necessarily represent the views of Nokia.
Since mobile
phones have a very heterogeneous set of properties and capabilities, content
adaptation and device independency are interesting to us. We have been looking
at different aspects of content adaptation since the year 2000, and collected
considerable experience in the area of server-side content adaptation [1]. We have examined server-side adaptation in
different forms: content selection based on metadata, content transcoding based
on media properties, and content transformation based on device independent
content. As an example of the latter, we participated in the European Union 5th
framework programme research project Consensus [2], researching a device
independent content format for mobile corporate applications. The learnings
from the Consensus project have been described earlier [3], but one of them is
worth repeating in this context: Content selection is a necessary ingredient
for device-independent authoring. We should perhaps also add a corollary:
Metadata is required for content selection, and therefore metadata is also a
necessary ingredient for device-independent authoring.
Currently
adaptation of content for mobile phones is performed mostly on server side
(small-screen browsing modes in latest browsers are a notable exception). This
has been a natural choice since the capabilities of mobile phones have until
only very recently been insufficient for client-side content adaptation.
However, the need for the server-side adaptation infrastructure stifles the
development of mobile content. An additional problem is the maintance overhead
of such infrastructure: the terminal profile databases must be kept up to date.
So far attempts to get rid of profile database maintenance, e.g. by including a
profile URL in request headers (static UAProf [4]), have not been very
successful.
Another problem
related to server-side adaptation is that sometimes the user would like to
store or forward the content for viewing later on another device. Since the
content has already been adapted for the user’s current device, it will not be
suitable for viewing on other devices.
Client-side
adaptation avoids these problems by not requiring the server to include
adaptation functionality. As a result, there is no need to transmit a
description of client properties to the server, and the server performs no content
modifications. Another advantage of client-side adaptation is that the client
knows its own properties intimately. It can therefore better guide adaptation
than a server-side system which has to rely on standardized profiles that may
be erroneous or out-of-date. A client-side adaptation system can also more
easily forward decision-making to the user when the right decision is not
obvious – for example, it can ask the user whether he wants a shortened,
high-usability or a full, low-usability version of the content.
As mobile phones
are becoming more capable, performing the content adaptation on the mobile
phone itself is rapidly becoming feasible. If there were a standard way to
describe different available versions of content, a phone could use such metadata
to decide which variant of the content to get from the server. XHTML 2.0
already has a basic content selection mechanism in place, in the form of nested
object elements. Currently the object element only supports selection by media
type, but it would be easy to extend it towards arbitrary selection conditions.
Another option would be a new XML specification meant especially for content
selection, such as the Content Selection for Device Independence Working Draft
recently published by the W3C Device Independence working group [5].
The two problems
where the client-side content selection approach seems most promising are size
and format. Even with the deployment of 3rd generation mobile networks, the
size of media content such as video clips remains a problem in the terms of
download speed, even though storage is becoming a non-issue. Similarly, even
though the capabilities of mobile devices are rapidly increasing, the number of
possible content formats is huge and keeps growing. Since many formats are
proprietary and support for them depends on third parties, we must assume that
format support will be a problem for mobile phones for a long time to come.
Presentation is
also a problem, although perhaps not as serious – a presentation meant for a
different environment usually does not
make the content incompatible with the terminal, merely hard to use (assuming
that all terminals understand the presentation markup). While media content has
issues with size and format, presentation of content is by its nature geared
towards a certain type of terminal. Therefore there should to be alternative
presentations to support different terminals. This could be e.g. in the form of
alternative CSS files, or one CSS file which contains alternative settings.
There are some
important requirements for the metadata if it is to enable the client-side
content adaptation we envisage. Firstly, it must be simple, so that use
of the metadata can become widespread. History shows that simplicity is more
important that expressiveness – authors have to be able to easily adopt and
understand the use of metadata. Secondly, it must be compact so that
even massive amounts of metadata will cause only neglible overhead. Thirdly, it
must be sufficient for common applications, and extendable – in a
well-defined manner – by special applications.
And finally, it
must be efficient – if it is slower or more difficult to interpret the
metadata than it would be to analyze the media content itself, there is no
point in it.
References
[1] Coulombe S, Koskimies O., Grassel G.: Content Adaptation for the
Mobile Internet, chapter 7 in the book ”Content Networking in the Mobile
Internet”, Wiley-Interscience, ISBN 0471466182, September 2004.
[2] Consensus project, EU 5th framework programme, http://www.consensus-online.org/.
[3] Koskimies O., Wasmund M., Wolkerstorfer P. and Ziegert T.: Practical
experiences with device independent authoring concepts. Advanced Visual
Interfaces 2004 Conference, Workshop on XML-based High-Level User Interface
Descriptions, May 25th 2004.
[4] Open Mobile Alliance: OMA UAProf 2.0 Specification, http://www.openmobilealliance.org/release_program/enabler_releases.html#UAProf,
May 20th 2003.
[5] W3C Device Independence Working Group: Content Selection for Device
Independence, W3C Working Draft, http://w3.org/TR/cselection/,
June 11th 2004.