|
From: Artyom B. <ar...@z2...> - 2003-03-20 10:31:43
|
On Thu, Mar 20, 2003 at 11:01:03AM +0100, Simon de Bakker wrote: > > but compound objects (or components, or maybe we'll call them servlets? > > because they run on server. then their GUIs will be applets) aren't the > > only way to hide details. the "intelligent pipeline construction" is > > another. this technique can take care of inserting ffmpeg-out / > > ffmpeg-in objects on both ends of distributed connector if the types of > > the properties being connected are supported (or suggest) using > > ffmpeg-stuff. > > So you get intelligent/flexible compound objects (servlets) that are > able to change in/out types according to connection. That's i think a > powerfull approach for hiding details.. > in the mean time i thought the following: CORBA objects should be the "servlets", they are higher level then the underlying simca_object_t based ones. they are, i guess, what i used to call components, except for GUI part (GUI is postponed to better times). i think on CORBA level we will need another interface - Link. it will represent either simple out->in connection between objects or the intelligent connection to remote object. in the second case Link object will hide actual network streaming mechanizm and in any case havig such an object will allow uniform representation of any connection - local or remote. i'm still doubting if the simca_object_t and friends should know about CORBA or not. i prefer them to not bother, but i don't know yet if that's possible. BTW, i've commited first .idl files to cvs, they may be oversimplified at the moment but at least we can start experimenting with this approach. -- Artem Baguinski: <ar...@v2...> <http://www.artm.org/> V2_lab: <http://lab.v2.nl/> <http://www.v2.nl/> |