[sv-bc] possible compromise on 578

We reached an impasse yesterday on my proposal, over the issue of allowing
implementations to treat ports with unspecified port kind as nets instead of
variables.  Would anyone change their opinion if this were only allowed for
input ports?  This might address most of the concerns expressed yesterday:

1. It would eliminate the condition "if there are no procedural assignments
   to the port".  For input ports, these are never legal anyway.
2. It would eliminate the most visible effect of seeing the locally driven
   value versus the external one.  An input port will not be locally driven.
3. On the PLI side, this might actually make it easier to use legacy PLI
   code, which expects only nets as input port loconns.
   
The relevant sentence would be changed to something like:

"However, if an input port has a data type supported for nets, an
implementation can default the port kind to a net of the default net
type instead."

Steven Sharp
sharp@cadence.com