My 2c

Hi

I very much agree with many of the comments put forward today. On the
other hand I think it is essential to (at least) interface to existing
standard available. OROCOS should not reimplement everything from
first principles as that simply is too much work and in addition it
might be easier to exploit exisiting standards. There are thus a
number of choices for *glue* that we ought to consider and on top of
these we ought to have well defined "interfaces" from which a variety
of implementations can be provided using different kinds of
*glue*. For that the design is critical. Just as an operating system
is an abstraction over a rich set of hardware components we need to
decide/design the abstractions that allow us to use the best (for a
given situation) underlying system for implementation. Thus we should
not standardise on the equivalent of IDE or SCSI interfaces but
provide the abstractions that allow us to use any of them. 

Igonring middleware and doing everthing on our own is in my view not a
good idea, as other have spent significnat resources developing
methods for communication, ... and it is not clear to me that we have
better expertise in all of these areas, we shoudl build on the
sholders of other giants rather than always starting from
scratch. This does not imly that we do not need generalised
interfaces/abstractions that gives flexibility in the underlying
system.

As to XML. I agree with David that we should not use XML as a run-time
messaging format, this is too inefficient, but for storage of
information and configurations it might be one viable alternative. Even
if writing XML parsers is a pain.

As to a single joint messaging / data structure format that is
centralized. I am not sure that this is a good idea. Initially I think
we ought to consider the set of libraries and services we would want
to see in a minimum system and consider how various systems can be
built from such a basic design. Initially I htink we should focus on
the basis and potentially a number of libraries. Across libraries they
is a need for standarisation of objects/representation, but initially
it might be wise to deligate this to sub groups that focus on
different aspects. A single unified framework from the start might be
difficult. 

I am happy with beer and wine!

--Henrik