I see. THe examples with the Task model that you give are realistic. They
illustrate the use of subclasses well.
Given the desire to use subclasses, we end up with the current restriction.
Correct me if I am wrong. If you want introduce an object before you give the
full definition, you at least have to specify its class.
I understand that when we have an interactive environment we will have to fix
this, because developers might want to change the class of a task that they
already created. In the current setup, they would have to go edit the place
where the task is introduced, and reload that portion of the MTF file. Not
optimal, but ok for the interim.
I agree that we should have some warning in the code and the documentation to
warn developers about the curren shortcoming.
I also pose the following question to Ewald, Pablo and you regarding the
object system. Would it be hard to change the implementation of objects so
that they have a dummy header that points to the actual object, so that one
can change the class without destroying the header object. If this is a
simple change to make, then we should consider making it.
Lets discuss this proposed change in our conf call later today. Please look
into this so that we are prepared with the info to make a decision.
Pedro Szekely
USC/ISI, 4676 Admiralty Way, Marina del Rey, CA 90292
Phone: 310/822-1511, Fax: 310/823-6714
URL: http://www.isi.edu/isd/szekely.html