See evaluations for the evaluation criteria.
The ARexx commands to mg are largely the commands of mg, which are well-defined. There are some arexx-specific extensions that are not clearly defined as yet.
The ARexx command set is the mg command set, which is defined to include everything the user can do - except for most Intuition manipulations. There are some Intuition facilities, but not enough to count as complete.
Since the ARexx commands in mg are the mg commands, they are consiste with mg, and themselves. However, they aren't very consistent with ARexx. The commands aren't Rexx tokens, but can be read as expressions. The textual data is interpreted in the manner mg uses, not the manner ARexx uses. However, the interface makes up for that by providing textual data via RESULT and the RVI for complex data, and using the return codes from commands to provide usefull information to the script.
The command set is consistent with mg's command set. This leads to one command per line, and long, descriptive names. This leads to commands that are easy to read, and debug. That the commands aren't Rexx tokens makes writing the commands somewhat more painfull than one would wish, as they must be quoted. Some would call that a feature. However, the textual arguments to the commands uses it's own strange quoting mechanism, and the interactions between that and Rexx make the text difficult to read, write and debug.
On to the thinker evaluation, back to the MFF+ evaluation, or up to evalutions.
Mike W. Meyer