PN Standardisation: A Survey

Petri Nets formalism requires standardisation to facilitate the work of researchers in this field and to enable the data exchange between different Petri Nets tools through a common format. Following this, a three-part International Standard (ISO/IEC 15909) has been developed. Part 1 is devoted to terms and definitions for Place/Transition Nets and High-Level Petri Nets. It is now completed (published as a standard) but will include an addendum on Symmetric Nets. Part 2 aims at providing a transfer format for High-level Petri Nets, called PNML, based on XML. Work on part 3 which deals with extensions has not started yet. In this paper the first two parts of the standard are presented. Then, to support part 2, an implementation of PNML, through an API framework to be integrated into Petri Net tools, is proposed. It allows for the translation of any Petri Net, designed by a given tool in a dedicated format, into PNML. 1 The Challenge of PN Standardisation Petri Nets [4,8,26,28] are a mathematically defined formalism and may thus be used to provide unambiguous specifications and descriptions of applications. They are especially dedicated to specify and design discrete event systems and this technique is particularly suited to parallel and distributed systems development as it supports concurrency. The technique allows for specification of systems at a level which is independent of the implementation choices (i.e., by software, hardware — electronic and/or mechanical — or humans, or a combination of these) and has been widely used to describe telecommunication systems, protocols, microprocessor architectures,... since their invention in 1962. They also constitute an executable technique, allowing specification prototypes to be developed to test ideas at the earliest and cheapest opportunity. Specifications written in the technique may be subject to analysis methods to prove properties about the specifications, before implementation commences, thus saving on E. Najm et al. (Eds.): FORTE 2006, LNCS 4229, pp. 307–322, 2006. c © IFIP International Federation for Information Processing 2006 308 L. Hillah et al. testing and maintenance time and providing a high confidence in the quality of the product to be developed. However these analysis methods are efficient only if they are supported by tools: for example, CPN-AMI [25], GreatSPN [9], PEP [13], CPNtool [22], automate the analysis process. A problem with Petri nets is the explosion of the huge number of elements when described in their graphical form, for specification of complex systems. High-level Petri Nets [17] were developed to overcome this problem by introducing higher-level concepts, such as the use of complex structured data carried by tokens, and using algebraic expressions to annotate net elements. The use of high-level concepts within this Petri net framework is analogous to the use of those in high-level programming languages (as opposed to assembly languages). In the Petri nets community the term High-level net is generally used to refer to nets using such concepts. Two of the early forms of high-level nets are Predicate-Transition Nets [12] and Coloured Petri Nets [16], first introduced in 1979 and further developed during the 1980s. Most of nowadays high-level nets build on these. They also use some of the notions developed for Algebraic Petri Nets [29], first introduced in the mid-1980s. Furthermore, there are many different variants of Petri nets. Extensions of the technique, including time, stochastic features, capacities, and hierarchies as well as special Petri net types exist in the literature (see [2,8]...). Standardisation of the technique has been seen as an opportunity to obtain a better organisation of the work in the Petri Net community. It has several issues: – to enable the stakeholder — researchers, as well as engineers using Petri nets — to use the same terminology; – to develop future extensions on a stable common basis, e.g., P/T nets or High-level nets; – to provide a reference implementation that will facilitate the data exchange between different Petri nets tools through a common format. The purpose of this paper is to present the PN standard, referenced under ISO/IEC 15909, as well as related work. First, the standard, which is organised in three different parts, is described. Then the current status of the work is given, followed by an implementation which is expected to prove very useful for the PN community. 2 The Structure of the Standard The PN standard has been designed into three independent parts in order to enable flexibility of the standardisation process. Part 1 [15] provides the mathematical definitions of High-level Petri Nets, called the semantic model, the graphical form of the technique, known as Highlevel Petri Net Graphs (HLPNGs), and its mapping to the semantic model. Part 1 also introduces some common notational conventions for HLPNGs. PN Standardisation: A Survey 309 Part 2 [18] [19] of this international standard defines a transfer format in order to support the exchange of High-level Petri Nets among different tools. This format is called the Petri Net Markup Language (PNML). Since there are many different versions of Petri nets in addition to High-level Petri Nets, this standard defines the core concepts of all Petri Net types along with an XML syntax, which can be used for exchanging any kind of Petri Net. Based on this PNML core model, part 2 also aims at defining the transfer syntax for the two versions of Petri Nets that are already defined in part 1 of this International Standard, Place/Transition Nets and High-level Petri Nets. An addendum to Part 1 [20] of this standard introduces Symmetric Nets, formerly known as Well-Formed nets [6], as a subclass of High-level Petri Nets, which uses a restricted set of algebraic operators and allows for good analysis possibilities. Part 3 is devoted to the standardisation of Petri nets extensions, including hierarchies, time and stochastic features. These extensions will be built upon extensions of the core model. They require a stable version of the core model to be available. This is not the current situation at this stage. Hence, only parts 1 and 2 are presented below. The standardisation process is quite long and relatively complex. A standard must be built in order to be stable enough to be used by the people involved. It is developed within a schedule which should not exceed three years and is subject to revision every five years. More information on the rules can be found at [10]. Part 1 obtained the status of International Standard in december 2004. The addendum has been proposed by France and has currently the level of Working Draft (stage 20.60 in the ISO nomenclature). At this level, the possibility is offered to the community to contribute. When the standard has reached the step forward, the Committee Draft level, it gains restricted aceess, with rights reserved to ISO experts only. Part 2 has today the same status as the addendum. Work on part 3 has not started yet, as it requires a stable version of the PNML core model to be available. As a consequence, the work on this part will start as soon as PNML is standardised.