← Back to team overview

opencog-dev team mailing list archive

Re: libatomspace branch

 

On Thu, 2008-07-24 at 10:54 +1200, Joel Pitt wrote:

> Nice one!
> 
> I'm assuming this'll make it easier to also have MindAgents (or groups
> of them) as separate dynamic libs which would be cool. Or is that what
> you meant by allowing the CogServer to be extended?
> 
> J

Not really. These changes turn opencog into an library/framework, but
the MindAgents are still static. Nothing prevents you from creating
agents as a static library, though; the inconvenience is that you'll
also have to extend the cogserver to plug in your agents... :-(

On the bright side, I already have a separate set of patches that adds
support for generic-dynamically-loadable modules to the cogserver. I've
finished the basic design last week and it works beautifully! :-)

Once both changes are merged, we'll have the best of both worlds:
dynamically loadable modules which can be built independently of opencog
on it's own tree.

Unfortunately the dynamic-modules merge will have to wait a bit. It was
developed on top of the 'threading' branch -- so I'll have to backport
it -- and worse, merging these changes to the 'libatomspace' branch will
make the current opencog <=> petaverse integration even harder, which I
believe is not really desirable right now.

--
Gustavo




Follow ups

References