[Data-modeling] Processes

[Jeff Prucher]

----- "Iain Sproat" <iainsproat at gmail.com> wrote:

> 
> Firstly, there is a Process type, e.g. Nitrogen fixation , which has
> some Input and Output properties (these would be typically expect a
> type of material, but to allow for flexibility I'm only going to
> expect a type of Topic)
> A process can be composed of other component processes (for the
> Nitrogen Cycle these would be Nitrogen fixation , decomposition ,
> Nitrification etc.). Parent and component processes are linked with a
> Process stage CVT .


I think it would be better to create a type or types for these: Process Input and Process Output, maybe. It's almost always better to create a type than to use /common/topic. For one thing, you can reverse the properties -- this would allow you to easily tell if something is part of more than one process; in some cases, anyway, I'm sure that would be useful or interesting to know.  For another thing, if a new topic is created via a property that expects Topic, that topic will appear to have no data associated with it (in the client, at the very least), which could very easily lead to it being deleted on the grounds that it's not being used. A lack of type/property information also makes it extremely to know what a topic represents (unless someone's put in a description), which can lead to either duplication or inappropriate co-typing.  (Sorry for the long-winded explanation!)

Jeff