Comments PR-mmi-arch-20120814

Sebastian,

Thank you for your detailed and careful comments.  You make a number of
good points, but it is unfortunately too late to make substantial
changes to the specification unless serious difficulties are found with
its implementability  or with the interoperability of implementations of
it.  We will fix the broken link to the Galaxy specification,

but more substantial editorial changes are not possible. 

 

To address a few of your comments in more detail, we would  first
observe that a specification is different from a  tutorial, and that
many W3C specifications have even less of an overview than ours does.
Thus even though you are correct that a more thorough introduction might
help the reader, such detail is not expected in a W3C specification.   

 

On the issue of the nature of life cycle events, we have deliberately
made their definition quite generic.  Since these events are sent as
individual messages, we do in some sense assume that modality components
are 'discrete' - or, more precisely - that their output can be
discretized.   For example, a continuous gesture interface could be used
within our framework if its output could be mapped to a series of life
cycle events or a streamed output  where small pieces of streams or
files are requested  and spliced together by Modality Components in an
interaction cycle coordinated with the life cycle events.

 

On the issue of the MVC paradigm, we refer to is as 'recent' because  it
has attracted a lot of attention in the context of web interface design.
We also think that our architecture would seem much closer to MVC if we
separated out the data model  (in our current definition 'M' and 'C' get
merged into the Interaction Manager.)   

 

On the choice of 'context' rather than 'session', we chose the former
term after a fair amount of discussion because it seemed fairly neutral
in a web context, while  'session' has a lot of pre-existing
associations.  Moreover, the context of interaction handled by the
Interaction Manager is different from the multimodal "session" as
defined in the  MMI Framework. The MMI Architecture recommendation
inherits from the MMI Runtime Framework a multimodal "session" that can
be joined by users and transferred between modalities. In contrast, the
notion of interaction context is placed at a different level of
abstraction than this multimodal "session" proposed by the MMI Framework
which concerns the environmental, state and system data.

 

Finally on the choice of HTTP as an example protocol, we do not mean to
imply that it is better suited than other protocols.  A lot of our work
is driven by the interests of individual working group members.  In this
case, one member of the group had built a prototype system using HTTP,
so he wrote it up and we included it in the specification.  

 

Thank you again for your time and close reading of the specification.
If we do not hear from you within a week, we will assume that you find
our explanations satisfactory. 

 

Jim Barnett