The separation of devices from phenomena is an important
observation and
that yields a better model of interaction
and leads to a more flexible architecture.
The publish-subscribe model of event notification
is a powerful concept that is poised to play an ever larger role
in both interaction and distributed computing.
(Notice that the signals/slots model in Qt is a move in this direction.)
The paper explicitly addresses and provides specific criteria
for getting sufficiently fast action-perception feedback,
a factor which seems often neglected in interaction design.
WEAKNESSES
Papers that give broad visions about exotic new forms of input
always make me wonder about the intended audience.
Who will bother to spend the money, provide the space,
and give the attention demanded by all of these devices?
More hardware devices and software complexity
means more ways to break down and more to maintain.
Because it now depends on all the input channels,
does a multimodal input system "raise the bar" for entry?
That is, does it become inaccessible to many more people
with any one of various disabilities?
Existing systems that use one channel are inaccessible
to those with a particular disability;
in some cases they can then adapt
by finding an alternative for that one input channel.
Do multimodal systems make this adaptation impossibly difficult?
The design vision described by the paper is likely
to break many user assumptions about coupling and independent systems.
For example, the PointRight example fundamentally changes
the way pointing works on multiple devices,
in a way that most users would not expect.
The assumption that a single physically isolated object
is operationally independent no longer holds.
As the number of involved components increases,
the user model must also become more complex.
zesty
ca)