Considered as a research project in SLP, considerable work needs to be done, including a specification of the semantics of WOM. It is likely that consideration of multiple viewpoints could help in designing a notation for classes which have more than one way of displaying themselves, and that some notions from Goguen's work on hierarchical systems [Goguen 1972] could be useful in providing a more suitable way to decribe forms which create sets of pages. The idea of nested time co-ordinate systems may help here. However, this work demonstrates that SLP can be used to design a practical and simple programming language.
As a practical tool for Web programming, Web-O-Matic is already useful, and I believe it to be a considerable advance on the CGI tools already available. Many of these are no more than collections of routines in Perl and similar scripting languages. There are some other object-oriented tools for Web programming, NeXT's WebObjects and the W3Kit amongst them, but these do not share Web-O-Matic's conceptual simplicity and conciseness. On the positive side, they have been in existence for longer, and provide a greater range of interface widgets and so on than does Web-O-Matic.
Although Object Rexx's long-lived objects made our run-time system simpler to implement, we fear that the speed and robustness of the generated code may be limited by the inefficiencies and bugs which we have discovered in Object Rexx. This product appears to receive less whole-hearted support from IBM than we would wish: they have been very reluctant to state how much they support Object Rexx now and in the future.
We are currently testing Web-O-Matic by connecting the IFS' Tax and Benefit model, TAXBEN, to the Web. Our recently released model Be Your Own Chancellor 1996 is built around this, but uses MDDL rather than Web-O-Matic. We hope that by the start of November, we shall have a more extensive version up which does use Web-O-Matic.
The current version of Web-O-Matic does not implement equations in constraints: the equation parser and part of the translator are written, but some of the dependency analysis and other code-generation work has yet to be done. As a temporary fix, I have implemented code inserts:
<TextField t1>
<Text t2>
<OnSubmit rexx>
{
if self~getT > 0 then
$t2~setValue( $t1~getValue )
}
</OnSubmit>
Here, the curly brackets delimit Object Rexx code, shielding the parser from the need to distinguish Rexx comparison operators from HTML brackets. References to instances are marked by dollar symbols, and the translator converts these into Rexx code to access the instance. By being careful about the kind of code one writes, it is still possible to write constraints cleanly, provided one avoids dependency loops.