SDL-2010: Background, Rationale, and Survey

This invited paper concerns a revised version of the ITU-T Specification and Description Language standard, which is scheduled to be consented for approval by ITU-T during 2011. In this document and ongoing ITU-T work, the revised version is called SDL-2010. The current standardized (or in ITU-T terminology Recommended) version at the time of initially writing this paper (April 2011) was called SDL-2000. The paper gives some historical background on the development of the language. The paper includes rationale for the update of the language and the revised organization of the language standard. After the history, there is a description of the new organization followed by some details of the changed feature set of a revised version SDL-2010 compared with SDL-2000. The paper concludes with a snapshot of the status of the SDL-2010 standard. The purpose of this paper is to provide an overview introduction to the revision of the ITU-T Specification and Description Language called SDL-2010. The Specification and Description Language has a long history as a published standard going back to 1976 [1], but other than corrections and some re-organization of material in 2002 [2], the language standard has remained stable since the publication of SDL-2000 in 1999 [3]. Considering that previously the language had been revised in 1980, 1984, 1988, 1992 and an addendum to the 1992 version in 1996 [4–8], this is the longest period of stability in the evolution of the language, as can be seen in Fig. 1. As explained in Sect. 1.1, this stability is to some extent due to changed market circumstances in the last decade. One incentive for a revised version of the language is to remove some differences between the language standard and implementations provided by tools: in particular some of the aspects of data introduced in SDL-2000 that have not been implemented (see Sect. 1.2), nesting of diagrams (see Sect. 1.3), and features without SDL-2000 semantics from the United Modeling Language (see Sect. 1.4). The resulting feature set is outlined in Sect. 1.5. The organization of the standard documents is described in Sect. 2, followed by more detailed description in Sect. 3 of some features added or deleted. Finally the paper closes with the status at the time of writing in sect.4. 1 The content of this paper is not entirely new because it reports work that has been in progress for some time in the domain of ITU-T standardization, but although some material has been made available as ITU-T temporary documents and some material presented at SAM-2010, it has not been widely published. Fig. 1. Specification and Description Language publication dates 1 Background for the development of SDL-2010 SDL-2000 was completed in 1999 with versions of the ITU-T Recommendations Z.100, Z.105, Z.107 and Z.109. The main document was Z.100 [3] with Z.105 and Z.107 [9, 10] covering use of ASN.1 [11] with the SDL-2000, and Z.109 [12] covering use with the Unified Modeling Language (UML)[13]. A revised Common Interchange Format for SDL-2000 in ITU-T Recommendation Z.106 [14] supplemented these in the year 2000. Since then there have been some minor updates to these Recommendations. The text was reorganized in 2002 so that Z.100 [2] describes the graphical language, and the parts of the textual (SDL/PR phrase representation) that are alternatives to graphical representation (SDL/GR) were moved to the interchange format in Z.106[15]. At the same time a number of corrections and a few minor changes were made. In 2003 an Amendment [16] was issued to incorporate two new Annexes B and C that concern backwards compatibility and conformance to the standard. In 2007 Z.100 was republished incorporating all agreed amendments [17]. The changes have been minor: either re-organization or correction of flaws in the 1999 version, so that essentially SDL-2000 has not changed and has remained stable. 1.1 Status of SDL-2000 When SDL-2000 was being developed, right up until the ITU-T meeting at which it was approved there were two sizable software organizations that were promising to produce tools to support SDL-2000 in 2000 or 2001: Telelogic and Verilog. A merger of these two organizations in Telelogic was announced before the end of 1999, so that some competition was removed in the tool market. Although these commercial tools already supported some of the features of SDL-2000 by 2000, it is now unlikely that there will ever be a tool that approaches full support of SDL-2000 in its final form [15, 17–19]. Even the tool that best supported SDL-92, Cinderella, reached a position by 2007 when it would probably never offer full SDL-2000 support, because it has a smaller (at least in value terms) share of the ITU Specification and Description Language tool market and had to offer compatibility with Telelogic as the market leader at that time. Cinderella collaborated with Humboldt University that previously had not entered into the commercial tool market. The Humboldt SDL-tool implemented many of the features of SDL-2000 on a trial basis to test the feasibility of various ideas indeed some features such as nested packages were implemented specifically to support feature requests promoted by Humboldt for OMG related work. In 2003 SOLINET announced the SAFIRE tool set, claiming that it is based on Z.100. In late 2004 PragmaDev, which previously supported a dialect called SDL-RT announced support also of Z.100. All four organizations (Cinderella, PragmaDev, SOLINET/SAFIRE, Telelogic) had commercial tools available in May 2006, though SOLINET/SAFIRE had ceased to be involved in ITU or SDLForum activities and the future of the language. By November 2008 all Telelogic products and services had become part of the IBM Rational Software portfolio, and the main tool vendors were IBM, PragmaDev and Cinderella (probably in order of market value at that time). At the time of writing all three of these vendors still offered Specification and Description Language tools. Since 1999 the general market perception has developed. In 1999, part of the rationale for developing Z.109 as a UML profile for the ITU Specification and Description Language was because UML was perceived as a major competitor to the ITU language. Within the telecommunications industry some organizations were divided internally between those that favoured the ITU Specification and Description Language and fans of UML. A decade later the perspective is quite different, because the issue is not seen as whether to use UML or SDL-2000 (and other ITU languages), but how to use these together. In retrospect it would be easy to say this was always the way it was seen but to be truthful this was not the case especially in 1997 and 1998 when SDL-2000 was being formulated. However, it is now clear that the state machine specification part of UML 2.1.2 is not really a complete language in itself, because of the semantic and syntactic variations that are allowed, and that to make its use practical an (implicit or explicit) profile for UML has to be used. The revised Z.109 profile of 2007 [20] is geared to the needs of the telecommunications industry by mapping UML 2.1.2 [13] onto the more precise (and therefore more practical) Z.100 semantics and (where UML 2.1.2 gives notation options or no specific notation or no notation) binding to the Z.100 syntax. It is not by accident that the situation has been reached today where UML and ITU System Design Languages are seen as complementary rather than competing. Between 1999 and 2004 there was a significant involvement of ITU System Design Language experts in the ongoing development of UML, in particular for UML 2.0. The ITU languages have the good features of being well-defined and having action semantics that ensure specific behaviours. UML is good at object modelling and has proven to be a success at providing a framework for using different languages together a feature that the ITU languages (for historical reasons) lack. Rather than defining new precise action languages for UML, or adding a framework scheme and object modelling to the ITU System Design Languages, the sensible way forward from a telecommunications system engineering point of view is to combine these features of both approaches. It was therefore not a surprise to see the industry use tools that combine UML with the SDL-2000 semantic engine. This was the perspective of several major telecommunications manufacturers, and therefore the general direction of industry. However, the situation with SDL-2000 after over a decade is unsatisfactory for all parties. Despite the 1996-1999 intention to ensure the language standard and tool support should be closely aligned (of course, ideally the same), this has not been the reality right through to the start of 2011. The language available to users is effectively SDL-92 with the 1996 addendum plus some of the features of SDL-2000 (which features depending on which tool is used) and often using legacy syntax for data. It was to ensure users are still able to produce SDLmodels that are valid according to the standard that Annex B was added to Z.100 for SDL-2000, which allows the legacy syntax supported by tools. Not only was SDL-2000 not fully supported, but also as UML and other languages such as the ITU-T User Requirements Notation [21], become more commonly used, there have been changing expectations of the facilities offered by the Specification and Description Language. Some, such as UML-like syntax, do not seem to be required. Others, that are not included in SDL-2000 such as timer supervised states, seem to be desirable.