Re: [COMPONENTS] Gnome's OAF allows `Capability-based activation'...

On Sun, 9 Sep 2001, Peter Soetens wrote:

[...]
> What might be a bit strange is that you define a component as something
> "independent" (from what ?) but you still state that a component can depend
> on the context, under strict conditions.

"independent": it can run by itself, needing nothing else but the
operation system.

"depending on the context": experience has shown that _any_
specification is context-dependent; you can't write everything that is
absolutely necessary to know about the functionality of a component in
the interface specification. For example: ask five people to write the
interface to a component that calculates the Jacobian matrix of a
robot, and all five can come up with something different, while still
using the same specification. Why? Ever thought about specifying all
these things: are you using orthogonal reference frames? right-handed
frames? 6-vector or 4x4 representations of velocities? what is
the meaning of the translational velocity vector? what units are you
working with? do you place rotational or translational components on
top? what exactly are the three references you need to uniquely
specify a relative velocity? Does your internal algorithms are
singularity free? if not, what singularities do they have? ...


> I have found another definition (
> http://www.dis.uu.se/~zeykiz/Papers/CBSEreport.pdf ) which is more elaborated
> and that paper also discusses other interesting definitions like 'framework',
> 'interface', 'patterns',... and the relation between them.

Thanks for the link!

Herman