VETk: Virtual Environment Toolkit

Introduction

VETk `Virtual Environment Toolkit' is a programming toolkit for writing distributed or concurrent user applications, possibly incorporating user interface agents. Examples are web-based multimodal dialogue systems and multi-user virtual environments.

The toolkit provides:

• network- and browser-transparent communication facilities
• a relational database engine with publisher-subscriber facilities
• standard components for writing GUI and Web applications
• Several special languages:
  • Components are written in VCL (VETk Component Language), which provides a thin yet convenient wrapper around Java, making integration of Java classes with the framework easy. VCL components compile to Java applets.
  • Component configuration may be done in both HTML with embedded agents, and VETkScript, which is a Python-based language, incorporating database access operations in Python. VETkScript also enables command-line operations, which is useful for online debugging and even online editing of system state.
  • Constraints on the data in the shared database may be defined in VDC (VETk Data Constraint definitions). The language enables entity/relationship modelling, statecharts, and OCL-type constraint definitions. Diagrams and constraint checking code may be generated from VDC specifications.

With the current toolkit, small applications can be developed quite easily. Very simple multi-user applications only take about a page of HTML code, which runs in any Java-compliant Web browser. Some more complex applications are included as example programs in the toolkit.

Distribution

This software is distributed under the BSD license. It requires:

• Java 1.1 or higher
• SWI prolog (it has been tested with version 3.4.2).
• Jython 2.1a1 (or perhaps a newer version) patched with the indent hack. It's probably easiest just to get the patched Jar here.
• The XVCG graph layout tool (tested with version 1.3).
• The standard Unix tools C, Perl, Bison, Flex

• DOWNLOAD RELEASE 2.4
• ONLINE REFERENCE MANUAL

  Changes w.r.t. release 2.3.4:
  Features: Added VETk Data Constraint language. Improved WebSet example.
  Bugfixes: The engine was made JDK 1.1 compliant. The program now runs out of the box on FreeBSD.

• Older versions: Release 2.3.4 Release 2.3.3 Release 2.3.2 Release 2.3.1 Release 2.2

Gallery of specifications

• A natural-language description
• The data constraint definition (VDC)
• the code for each window (HTML or VETkScript)
• any startup scripts of standalone agents (VETkScript)
• if other than standard agents were used, the code for each agent. (VCL)

• Webset
• Mini VMC
• Simplechat (Screen grab)
• Shared Blackboard

Future plans

• The role of VCL should move towards that of a component interface language, providing descriptions of interfaces only without mixing in code. This should provide more language-independent interface descriptions, making the framework ready for multi-linguality. The interfaces should be connected to back-ends written in any language by means of a specifiable back-end interface. For example, we might specify a Java back-end, making a well-defined connection of a given VCL interface with Java method calls, and a Prolog back-end, making such a connection with execution of Prolog clauses. To implement this easily, this does require some form of reflection, for example, the Java back-end should be able to seek out methods in specified classes.
• OCL-type expressions should move to VETkScript, since they prove quite useful. State machines too could move to VETkScript, making them part of the system's description rather than of its behaviour constraints. This last idea was prompted by the need to be able to read out the state of a finite-state automaton in many cases.

Implement a method for combining synchronous and asynchronous publish/subscribe functionality, enabling parts of the application to operate synchronously. Actually, I've been thinking about this for a long time, and have identified many problems with such an approach. Most likely is the addition of a limited synchronous form of communication.

For a long while I've been thinking about a 3D library. To be interesting, it should be capable of animations. However, the easily available 3D facilities (VRML, Java 3D) have their technical weaknesses, probably making the development of such a library a lengthy and difficult enterprise. I've even been thinking of re-implementing VRML, which, from my research viewpoint, has some interesting features, including an interesting dynamical model. I think it's a pity that the development of the VRML language seems to have stopped, as the language might have been very nice with further improvements. Rather, development seems to be going into the wrong direction, with the movement towards X3D. X3D is only different from VRML in a shallow way, and it seems that the movement towards an XML syntax moves the role of VRML away from a programming language towards a data format. To start with, I think that the VRML syntax BNF can easily be improved, and the language should instead incorporate some extra practical features to make it more useable as a programming language, such as macros and numerical expressions. Here is my attempt at a VRML syntax, at least this one is LALR(1) and is also a bit smaller and more extensible than the VRML standard grammar.

Background information

Papers and presentations on VETk

• The paper
• Slides + notes black&white, for printing
• Slides + notes colour, for screen-reading

• Abstract
• Slides + notes, B/W, Slides + notes, colour.

Related work

• List of models and toolkits
• Software engineering concepts
  • Reflection
  • Configuration languages
  • Coordination models and shared dataspaces