Environmental changes drive the evolution of organizational software systems. i * allows to model continuous requirements on social and intentional aspects of organizational software systems. However, partially computer-enacted work ows are not well supported by current tools. In this paper, we propose i*-REST, an Open Source collection of RESTful Web services that are designed to manipulate existing i * models on-the-large. More speci c, we describe two services in detail. The rst one allows the creation, modi cation and storage of Strategic Dependency models. The second service embeds i * model visualizations in arbitrary Web pages. We have evaluated the services in a case study on learning forums. We introduced a work ow for the re nement of existing archetypical i * models of learning communities without human intervention. Maintenance and evolution of existing i * models will be much easier, in particular when more i*-REST services become available.
Organizations evolve as they require rapid adaptations to environmental changes.
Sustainable success of organizations is only possible by continuous support of
their systems [
Jarke et al. [
J-PRIM [
In this paper, we describe RESTful services that allow to maintain i * models
by receiving commands in the form of REST requests from other applications.
i*-REST provides a storage with versioning support where models are stored
in iStarML [
In the following, we give an overview of the i*-REST service architecture. After that, we discuss creation, editing and deleting models using REST requests. Later we point out the service possibilities to visualize, import and export i * models. We nish the paper with our case study, discussions and plans for the future.
i*-REST consists of two services. One of the services maintains i * models by creating, editing, deleting and storing them as iStarML les. The other service creates SVG les that can be viewed by any Web browser. The architecture of i*-REST (Figure 1) allows the services to be distributed in peer-to-peer systems. iStarML Visualizer
Service iStarML Model
Service
iStarML Visualization Servqiceueries Clients Clients
Web Connector
The Web connector handles REST requests from outside and converts them into method calls of the i*-REST services.
The XML database serves as an i * model repository. An eXist1 database allows the storage of XML les and their versions e ciently. 1 http://exist-db.org/exist/apps/homepage/index.html
The iStarML Model Service creates, modi es and retrieves i * strategic dependency models. An example of an iStarML le is given in Listing 1.
Listing 1: Initial model as an iStarML le 1 <istarml> 2 <diagram name="diagram"> 3 <actor id="1" name="Homework" type="actor"/> 4 <actor id="2" name="Teacher" type="actor"/> 5 </diagram> 6 </istarml>
The RESTful API of i*-REST allows to create and modify i * models using requests that are related to the iStarML syntax. The rst request (Table 1) results in the code from Listing 1. The goal dependency between both actors is created by requests 2-4 and will add the lines from Listing 2. These dependency links and the goal dependency are deleted by the request 5. For the complete list of supported REST requests an interested reader can visit the i * wiki page of i*-REST2.
Listing 2: An intentional element, its depender and dependee in iStarML format 3 4 5 6 7 8 <ielement id="3" name="evaluate" type="goal"> <dependency> <depender aref="1"/> <dependee aref="2"/> </dependency> </ielement>
The service keeps iStarML les always valid by rejecting invalid operations, e.g. adding an actor link to an intentional element is not possible. Moreover, the service handles changes in models intelligently, e.g. a node is deleted with all dependency links or an intentional element is deleted together with its dependency links.
The iStarML Visualization Service creates visual representations of i * models by converting iStarML les into SVGs that can be embedded in arbitrary Web pages. The graph representation of a model is generated using the yFiles3 library. It ensures a compact graph layouting and usage of colors for nodes and labels.
We create a Web page that allows user interactions with a model repository.
The search functionality is realized by utilizing XQuery [
The visualization representation can be navigated, similar to an online map, by dragging and zooming with a mouse or a keyboard. Comments speci ed inside an iStarML le are shown as tooltips on top of the SVG representation. A user can download a visualized model as a SVG le (export) and import a local iStarML le to store and visualize it. 2 http://istar.rwth-aachen.de/tiki-index.php?page=i%2A-REST 3 http://www.yworks.com/en/index.html GET Collection/model1 PUT Collection/model1/ielement /3?type=goal&name=evaluate PUT Collection/model1/ielement /3/depender/1 PUT Collection/model1/ielement /3/dependee/2
Description Returns the model stored in a model repository.
Creates an intentional element of type goal (ID=3) in the model.
Adds a dependency link between the actor (ID=1) as a depender and the intentional element (ID=3) in the model.
Adds a dependency link between the actor (ID=2) as a dependee and the intentional element (ID=3) in the model.
DELETE Collection/model1/ielement/3 Deletes the intentional element (ID=3) in the model.
Similar to organizations, learning communities have their actors, policies, rules
and tools that they have to adopt because of environmental changes. We model
learning communities using i *, where learning community environments and
their customers are i * actors and learning materials, goals, tools, and tasks are
dependencies between the actors. We present archetypical community models
in learning environments like the innovative community, the question-answer
community, and the expert community [
In our study we investigate two learning forums: one with 430K posts and
21K users over 9 years, the other with 200K posts and 25K users for more than
13 years. Both are learning forums: the rst prepares language learners for tests;
the second supports physicians in lifelong learning. We detect communities using
the Louvan algorithm [
A community model suits one of the archetypical i * model, e.g. innovative or
expert community models. After nding a suitable model for the community we
mine community goals [
Before we implemented the services of i*-REST in the community, we evaluated their usability. Seven developers simultaneously used the i*-REST services for creating and editing small models (<10 actors, <10 dependencies). Each participant worked with her own set of models. We created a Web page where users sent REST requests to create and modify i * models. The idea to use REST requests was well accepted, the generated visualizations and service response times were satisfactory and the developers preferred Web pages to native clients for i * model visualization. However, the last result should be precisely investigated, e.g. by comparing Web browser-based and desktop tools for i * modeling.
In this paper we introduce the light-weight technical contribution. The purpose of the contribution is the computer-enacted support of continuous requirements changes. Our implementation allows to maintain and store i * models and their storage using RESTful services. In our case study, we realize the work ow of modeling continuous requirements without human intervention. We use i*-REST for creating models based on continuous analysis of learning communities appearing in Web forums. Model visualization of Web-based learning communities are helpful for their stakeholders. They can analyze community situations, adopt communities to suit a successful pattern, and provide additional functionalities to support communities to achieve their goals.
On the example of learning forums' environment we introduce continuous
support of evolving communities. Applying i*REST services that analyze forum
communities we keep i * models of communities consistent with their evolution.
Stakeholders can learn from existing i * models and choose design alternatives
[
We will extend i*-REST by supporting Strategic Rationale models,
visualizing di erences between model versions, adding evolution links between user roles
[
The RESTful approach provides a strong basis for interoperability and exibility, where services and users can harness the full power of a growing service repository. Introducing i*-REST, we encourage the i * research community to contribute with other i * RESTful Web services. They may facilitate collaborative modeling and sharing of i* models between organizations and communities.
This work was supported by NRW State within the B-IT Research School and RWTH Aachen University.