Global description of the project

The specification and the implementation of the system necessary for the project has been given as an assignment to the Software Engineering class, as a semester project. What is presented here is thus the result of a collaborative effort and shouldn't be considered the work of just the author of this document.

WWW was chosen as a support system for this project as it allows for remote access and provides a highly portable display interface abstraction. This includes the Mosaic viewer, as the system we have implemented makes heavy use of forms as they are implemented since version 2.0 of Mosaic for X. This approach should make it easy for a student to use this on-line facility from any personal computer or workstation on campus, or even anywhere else in the Internet world.

Since the book was available in electronic form, putting it in hypertextual form was not much of a problem and will therefore not be discussed in this paper. The main problem that had to be solved was to monitor the execution of Pascal programs through WWW, taking into account the fact that the http protocol is stateless, that is, the server does not remember former queries when it processes a new one (see note).

This problem can be further decomposed into:

1. How to run a WWW-activated program step by step.
2. How to display the intermediate states of the program, including its data structures.
3. How to manage user interactions with the model to view or modify the content of variables or set breakpoints in the code.
4. How to minimize client-server traffic.

• First, we considered using a Pascal interpreter and modifying it to allow external monitoring of its states. This solution was abandoned, as we had available only a Pascal-S interpreter, that didn't include file I/O nor dynamic allocation (no pointer types). In other circumstances, using an interpreter could however be a viable solution. Actually, many http servers are already written in some interpreted language. For instance, the Plexus server is just a Perl script.

• We then considered rewriting every Pascal program as a finite state automaton that could start execution at any state and stop execution either at the next state, or at any state for which there would be a breakpoint. The program state and data would be transmitted to the WWW client, both as text to be displayed and, in a more compact form, as a URL hidden in a hyperlink associated with the Step and Continue push-buttons (see pseudo-code).

  As an alternative to transmitting the whole program data from the server to the client, than back to the server, this information could be stored on the server side, in a temporary file, and have the URL hold a reference to this file.

  This solution would however not be easy to implement for recursive algorithms, as it would imply the explicit management of a stack to store local variables and parameters.

• The solution we finally chose consists in having the shell script, invoked by the http server, spawn a child process to run a slightly modified version of the Pascal program. The modification consists in adding, between each original Pascal instruction, a call to a synchronization procedure. The synchronization procedure checks whether the execution of the Pascal program should continue, or be interrupted.

  Whenever there is a breakpoint on the current instruction, or when the user has selected the single step button, the current state is sent to the client viewer, and the synchronization procedure waits for a signal from the server-activated shell script to resume execution, with possibly new breakpoints and selected variables, based on the latest user request.

Screen design and user interface

Site Hosting: Bronco