IPE and SepView

As I said earlier, Jean-Claude Dulac's SepView user interface manager is similar in concept to IPE but far more general. Here are a few of the important differences between the two:

• SepView allows for a non-linear processing sequence. For instance, the user can input or output auxiliary datasets anywhere in the sequence, or split the sequence into two branches. IPE could handle the auxiliary datasets with small modifications, but branching would require a great deal of effort.
• In SepView the descriptions of program parameters are interpreted, while in IPE they are compiled into the Ipe executable. This means that in SepView the descriptions can be modified ``on the fly'' while in Ipe you would have to re-run Ipemake, and then re-compile everything.
• SepView allows the parameter descriptions to include some control logic. Parameters can be checked, the user can be prompted to supply missing or badly-specified parameters, and so on. IPE takes whatever the user inputs and passes it along to the program as is. Any error messages go to the screen, just as they would for batch execution.

Obviously, these differences all point in SepView's favor. But as one can imagine, SepView's complexity makes it more difficult to use. Constructing the parameter descriptions for even one program is a formidable task. IPE uses the very simple description scheme described above, and it is realistic to add a new program to IPE in the space of fifteen minutes.

My goal is to first make IPE a stable, useful product, supporting a large number of programs, and then gradually add features to it to make it more powerful. Undoubtedly some of the additions will be patterned after features of SepView, which I think is a really nice piece of work. I am particularly interested in allowing for a non-linear processing sequence. I am also interested in checking input parameters and ensuring that the processing sequence the user constructs is valid.