[Pipet Devel] NetInputs and NetOutputs

[Jean-Marc Valin]

> About DL support: We'll need commands like "create_link" and "destroy_link"
> ("link" is Overflow terminology, which we should use, and a "pipe" is a type
> of link).  As an example of why the UI needs to follow the commands of the DL
> (although that may seem backward), imagine the user entering a subnet and
> disconnecting a link.  The UI tells the DL......

I'd be tempted to use "Link" in the PL and "Pipe" for the BL. A link is
implemented through the getOutput() call and the exchenge is made in memory,
while a "Pipe" is implemented through the BL and uses CORBA/sockets/... 

> About iterations: The idea of using iterators with data passed around the
> Internet (not just within the Overflow system) is interesting.  Say that you
> want to pass some data to a remote processor in a loop until a condition is
> met.  But, where should the logic/controller for this reside?  Do we need to
> (re)implement this in the BL layer, or should the user make a PL subnet to
> handle this?  Thoughts?  Talk amongst yourselves :-)

Just like in Overflow a Network is a Node, and just like the BL can wrap
Overflow, we could have a PL node that "wraps" the BL (maybe by executing
another BL on the local machine).

	Jean-Marc

-- 
Jean-Marc Valin
Universite de Sherbrooke - Genie Electrique
valj01 at gel.usherb.ca