《XMILE towards an XML interchange language for system dynamics… .pdf》由会员分享，可在线阅读，更多相关《XMILE towards an XML interchange language for system dynamics… .pdf（9页珍藏版）》请在三一文库上搜索。
1、V. G. Diker and R. B. Allen: An Interchange Language for System Dynamics Models351 Abstract We propose that XML should be used as an interchange language for system dynamics models and we have designed and implemented a demonstration that we call XMILE. We believe that there will be many benefi ts f
2、or the system dynamics community in developing a full-scale interchange language. This paper discusses the motivation factors behind the idea of developing an interchange language for system dynamics models, and explains why XML is a good candidate framework. An initial implementation of XMILE is al
3、so presented along with initial fi ndings about its applicability to models built on various system dynamics software platforms. Copyright 2005 John Wiley see the Acknowledgements section for the origin of the term SMILE). The scheme can be used as an interchange standard to enable conversion of mod
4、els among system dynamics modeling and simula- tion software. XML (W3C-XML) has many advantages as a framework for an interchange standard: It is fl exible, widely accepted, and there are many tools to support its use. Motivation An interchange standard offers many advantages to individual modelers
5、and to the fi eld as a whole. Individuals will be able to view and run models built by others without the extra effort of manual conversion between software-specifi c Vedat G. Diker is Assistant Professor at the College of Information Studies at the University of MarylandCollege Park. He received hi
6、s PhD in Information Science from the State University of New York at Albany. Robert B. Allen is now at the College of Information Science and Technology of Drexel University, where he teaches courses on information retrieval and digital libraries. He taught at the University of Maryland from 1998 t
7、o 2004. Before that he was Senior Scientist in the Information Science Research Group at Bellcore. He received his PhD in Experimental Psychology from UCSD and joined Bell Laboratories in 1978. He was the Chair of the Publications Board of the ACM and was Editor in Chief of ACM Transactions on Infor
8、mation Systems for 10 years. System Dynamics Review Vol. 21, No. 4, (Winter 2005): 351359Received September 2004 Published online in Wiley InterScienceAccepted August 2005 (). DOI: 10.1002/sdr.321 Copyright 2005 John Wiley & Sons, Ltd. 351 XMILE: towards an XML interchange language for system dynami
9、cs models Vedat G. Dikera* and Robert B. Allenb NOTES AND INSIGHTS a College of Information Studies, University of Maryland, 4105 Hornbake Building, College Park, MD 20742, U.S.A. E-mail: dikerglue.umd.edu b College of Information Science and Technology, Drexel University, 3141 Chestnut Street, Phil
10、adelphia, PA 19104, U.S.A. E-mail: rbadrexel.edu * Correspondence to: Vedat G. Diker, College of Information Studies, University of Maryland, 4105 Hornbake Building, College Park, MD 20742-4345, USA. E-mail: dikerglue.umd.edu 352System Dynamics Review Volume 21 Number 4 Winter 2005 formats. Modelers
11、 will be able to reach a wider audience not only with written work based on models but also with running models. The fi eld of system dynamics as a whole will also benefi t from an interchange standard. Specifi c examples illustrate the potential benefi ts. Reinforcing peer review An important requi
12、rement for a scientifi c fi eld to stay “healthy” is the ample availability of venues for peer review. For a fi eld like system dynamics, which has the activity of building and analyzing models in its core, peer review of models is as crucial as peer review of papers and reports. There are many “pas
13、sive” ways of reviewing a model, including critically reading the equa- tions of the model and investigating the printed graphs of model behavior from simulation runs carried out by the modeler. However, an “active” approach that involves the actual simulation of the model by the reviewer is by and
14、large the most rigorous approach. A serious barrier for such an active approach occurs whenever a person reviews a model that is built in a software platform different than the one he/she uses. System dynamics software companies are generally very good in supplying free limited-edition versions, mod
15、el readers, or trial versions of their software. So, obtaining the “right” software to run the model is not an insurmountable barrier. However, locating, downloading and installing the “new” software can become a time-consuming and tedious task. Even more time- consuming is learning the new software
16、, if the reviewer is not familiar with it. The reviewer may simply give up at some point of the lengthy process and go back to “passive” reviewing methods (that is, if artifacts about the model that lend themselves to conventional reviewing methods such as papers do exist). A model interchange langu
17、age could foster a higher level of active peer review within the community by facilitating conversion of models between different software platforms. Lowering barriers to entry for valuable niche software The fi eld of system dynamics has always benefi ted from high-quality, highly accessible softwa
18、re. New tools for model development and analysis provide new ways of looking at our models and improve the speed and depth with which we understand the problems at hand and develop plausible solutions. System dynamics software companies constantly strive to incorporate new tools in their respective
19、software. However, some potentially valuable addi- tions to the system dynamics “toolset” cannot make their way into mainstream use if the inventor fails to convince one of the few major system dynamics software companies about the value of adding the new tool to their software as a feature. In such
20、 cases the inventors have the option of delivering the new tool as a stand-alone, “niche” software product. This comes with a number V. G. Diker and R. B. Allen: An Interchange Language for System Dynamics Models353 of challenges, and a really big challenge is to handle models built on other softwar
21、e platforms. Without effectively handling all such models, the new niche software cannot reach its full “market” potential. The lack of effectively handling different fi le formats is an impeding factor in such cases. A model interchange language would remove this barrier and make it easier for new,
22、 useful software tools to reach a wide audience. Fostering increased collaboration between different research segments that have developed around specifi c software Certain segments of the fi eld of system dynamics have very specifi c choices about modeling software platforms they use. As an example
23、, Stella (Richmond et al. 1987) has long been the choice of researchers and practitioners working on K-12 education. This means that the users of other system dynamics soft- ware might have problems studying the K-12 applications of system dynamics, or fi nding a wide audience for their K-12-related
24、 models built on software other than Stella. This kind of barrier can be observed in other application areas as well. An interchange standard, which will facilitate easy and accurate conversion of models between system dynamics software, will improve the connections between different research segmen
25、ts within the community. A standard for system dynamics models As system dynamics software platforms evolve, they involve a growing number of platform-specifi c built-in functions and specialized variable types. Conse- quently, system dynamics models include a growing number of specialized structure
26、s or equations that are specifi c to the software platforms they are built in. This brings about a situation where the “same” model is not (or in some cases cannot be) represented “the same” in two different platforms. More and more, we are building “Stella models,” “Vensim models,” “Powersim Models
27、”, rather than system dynamics models. While it is benefi cial to take advantage of specialized features offered by different software platforms, this trend should not take us to a point where the semantics of a system dynamics model is ultimately defi ned by a given software platform.1 An interchan
28、ge standard can help defi ne and maintain the core semantics of system dynamics models. This would help distinguish the platform-specifi c features of a model from the core features. Facilitating model conversion BETWEEN DIFFERENT MODELING SOFTWAREThere is no doubt that easy and simple model convers
29、ions between different modeling software would make life easier for modelers who need to migrate a model from one software application to another. However, a more indirect, but equally signifi cant consequence of 354System Dynamics Review Volume 21 Number 4 Winter 2005 simple and easy model conversi
30、ons would be the possibility of using two or more modeling software applications in parallel while working on a single model. This would bring about a whole new way of building, analyzing and presenting system dynamics models. Modelers would be able to use one appli- cation to build, another to anal
31、yze, and yet another to present their models without extra effort, thus making use of the best features of each application. The option of analyzing or testing a model on several applications, making use of application-specifi c analysis features, would provide a wider spectrum of insights about the
32、 model, and arguably lead to a more robust model. Modelers also would be free to choose the most suitable application to present their models, based on considerations such as audience characteristics and model size. BETWEEN DIFFERENT VERSIONS OF THE SAME MODELING SOFTWAREIn addition to prob- lems in
33、volved in converting models between different software applications, converting models between different versions of the same modeling software can sometimes be a challenge, too. Modelers who have needed to study or improve a model that had been built several years earlier are rather likely to have
34、experienced those challenges. Easy conversions between different ver- sions would not only save time and effort on the part of modelers when such a conversion is necessary, but also make archiving and reusing models much easier. There would be a certain level of assurance that a model archived today
35、 will be available for use many years later. And when the time comes to revisit and reuse an old model, there would be no major compatibility problems between the model and the new version of the software application. This, hopefully, would motivate individual modelers and research segments within t
36、he community to invest more time and energy in archiving and disseminating their models. Facilitating collaborative modeling Last but not least, a standard, XML-based representation of system dynamics models would provide opportunities to use versioning tools such as CVS (Vesperman 2003) and sub-ver
37、sion (Collins-Sussman et al. 2004) to version models. This would facilitate large-scale collaborative model-building endeavors. Open standards XMILE will not be a translation engine. It will be a platform-independent, stand-alone representation of a system dynamics model. Translation from one model
38、format to another will be realized by fi rst translating the model fi le to a XMILE representation and then from XMILE to the other model format. The two translation “engines” will be independent. In fact, we expect translation V. G. Diker and R. B. Allen: An Interchange Language for System Dynamics
39、 Models355 applications (to and from XMILE) to emerge independently, possibly through efforts of separate people or groups. We envision XMILE as an open standard. This will motivate individuals and groups to take the initiative and develop the needed translation applications and it will shift the bu
40、rden from the software companies to the community as a whole. Furthermore, any given software vendor or developer will only need to build two applicationsone to parse the native data format to XMILE and one to transform XMILE to the native data formatinstead of having to build separate applications
41、to translate to and from each of the other native model data formats. XMILEs payoff increases as the number of native formats increase. Of course, the number of translation applications needed to translate everything among N platforms is N * (N 1), and thus grows geometrically, while the total numbe
42、r of parsers and transformers between N platforms and XMILE is 2N, which grows arithmetically. XML and XMILE The key insight behind XML is that the structure should be cleanly separated from the display format. In XML, the structure is defi ned as a set of hierarch- ically organized tags. The struct
43、ure we developed is shown in Figure 1 and forms a type of XML DTD.2 An example of the tagging is shown in Figure 2. Goals of the implementation We employed a simple population growth model to show how it can be converted from several implementations (Vensim and Stella) to XMILE. We also show how the
44、 XMILE version can be re-created as a Vensim (Eberlein and File Model Variables Name Type Value Units Comments Control Name Value Units Comment Display Unit Equivalents Game Controls Data Fig. 1. Tag hierarchy for XMILE 356System Dynamics Review Volume 21 Number 4 Winter 2005 Peterson 1992) fi le an
45、d can also be used by tools such as Digest. Our example is intended more as an illustration of this activity than as a comprehensive solution. We were also able to convert from Stella to XMILE based on the Stella text fi le. It would have been better to employ the Stella binary fi le but we decided
46、not to spend time on the time-consuming job of analyzing the fi le structure at this stage. Thus, the “units” were not available and we had to use default units. One of the virtues of separating the content from the format is that an XML content representation can be displayed in a variety of presen
47、tation languages. For documents, this translation is usually accomplished with the XML Style Transformation language (XSLT). However, we felt that the current imple- mentation of the XSLT scripting language would not be fl exible enough for this special application. Thus, we developed a C program to
48、 translate XMILE to Vensim format to demonstrate that such translation is feasible. Birth Coefficient constant 0.01 Dmnl Births flow Birth Coefficient*Population People/Year Population stock INTEG ( Births-Deaths, Initial Population) StockFlowBirths-Deaths StockInitialValueInitial Population People
49、This is the total population at a given time. Fig. 2. Fragment of tagging for a population model, including representations for a constant, a fl ow, and a stock. V. G. Diker and R. B. Allen: An Interchange Language for System Dynamics Models357 XMILE to Digest An example of the utility of this work would be to facilitate importing models into tools such as Digest (Mojtahedzadeh et al. 2004). Digest identifi es the dominant structure that determines the behavior of a given variable in a system dynamics model. Currently, Digest imports most Stella