Ricardo Amador
| Professional | Research | Contact |
| Ricardo Amador
|
|
| Research : PhD Proposal |
|
PhD Proposal2006-10-16 WIDERWeb Integrated Development tools for Evolution and Reactivity
Contents ^Objectives *Our goal is to define a reactive rule-based abstract execution model (together with an appropriate foundational ontology) capable of dealing with the distributed, non-centralized and heterogeneous nature of the Web (taking also into consideration transactional requirements). We expect to take this definition to the extent required in order to create a prototype of an integrated development environment (IDE) that facilitates the adoption of such a complex model. Having as our final goal to capture and demonstrate the functionality needed in a development environment (and the support that must be built into the related execution model), in order to make Semantic Web application building, based on reactive rule-based models, viable (and as simple as possible). One might argue that it is too early to start researching IDE requirements, since concrete languages are not yet fully defined. But that´s exactly our point, in order to deal with language heterogeneity (that is unavoidable in the Semantic Web) one must define a foundational ontology (with an associated abstract execution model) that by capturing the common ground shared by all these reactive languages works as an umbrella under which layered extensions may be created. Such an umbrella unifies and embraces more specific proposals like the current ones, and at the same time provides the foundation to investigate higher level issues that, at least to some extent, should not be specific to particular languages (like integrated development tools requirements). Motivation and State of the Art *Current approaches to Semantic Web programming tend to be service centric. Most of today´s Web applications are built using service oriented, usually opaque, procedural programming models. Although valid, these approaches may lead to a tightly coupled architecture with high maintenance costs. In order to turn the Web into a Semantic Web, as envisaged in [8], full semantic integration among heterogeneous systems (sites) is needed. Alternative approaches are being considered with special emphasis on declarative rule-based approaches with reactive capabilities (where services also have a role to perform). Such alternatives may yield a much more adaptive network, even one capable of autonomous intelligent evolution. In the context of a dynamic and evolving Web, models based on reactive rules (viz. Event-Condition-Action - ECA - rules: on Event if Condition do Action) have been proposed as an adequate solution (e.g. [15]). We believe that acceptance of such reactive rule-based approaches may be hindered by two issues: failure to capture technological heterogeneity of the Web, and lack of development tools that will make (the inherent complexity of) such models usable/programmable in a multi-party distributed scenario. Reactive rule-based models have been investigated and used in other research fields, like Active Databases (ADBMS) or Multi-Agent Systems (MAS). Some research efforts are trying to decouple these reactive execution models (e.g. ECA) from the underlying technologies (e.g. ADBMS) and use them for the development of distributed Web applications. These efforts tend to focus on developing new languages, which, although unquestionably needed, tend to hide the fact that the Web (and Computer Science in general) is technically an heterogeneous environment where lots of languages coexist. We believe it is possible to define an abstract reactive execution model that, instead of enforcing a single language, fosters the use of the most adequate language for each case (either a known language or a new one). Such an abstract model actually extends the open-world assumption of the Semantic Web from the data level to the behaviour level. During the last two decades of the past century, models based on ECA rules have been profoundly investigated in the context of ADBMS [16]. According to [4], since 1998 the number of publications in the ADBMS field dropped substantially, and today most of the research groups previously involved in this field have focused their efforts in other areas of research. Despite this, adoption by the industry of the results obtained [17] has been quite limited. In fact, this slow rate of adoption was already predicted in [12] and [17], given the lack of results concerning: analysis and design methodologies, and integrated development tools. As pointed out in [5], models based on ECA rules have been investigated also as a tool for behavioral modelling of MAS. Although this is still an active field of research, areas related to Software Engineering have not yet produced significant results, particularly at heterogeneous (meso) and global (macro) scales [21], like the ones entailed by the Web, thus raising obstacles to the generalized adoption of these systems. In the last years, results towards adapting these reactive models into a framework for evolution in the Semantic Web are starting to appear, at a quite rapid pace. These results cover a broad range of research fields, viz.: definition of generic rule markup languages (e.g. [9]), definition of event algebras (e.g. [7]), declarative specification of queries (e.g. [19]), definition of (trans)action languages (e.g. [20]), models for ontology based language composition (e.g. [18]), rule based definition of access policies (e.g. [10]). Recently, the first proposals for full reactive rule-based languages are starting to appear (e.g. [11]). Yet, to the best of our knowledge, no work is being done in the field of integrated development environments (IDE) that might facilitate the generalized adoption of these languages, and none of these language proposals accounts for language heterogeneity wrt. rule sharing and reasoning about rules (as proposed by our current work [3], [14], [13], [2], [6], [1]). When trying to unbundle ECA rules from ADBMS, there is an additional issue that emerges: how to capture transactional behavior (including cascading reactions). ECA rules have some times been extended to Event-Condition-Action-Postcondition (ECAP) rules that deal not only with pre-conditions but also with post-conditions and effects. We believe that ECAP rules may shed some light on the issue of transactional behavior when trying to define a reactive execution model with transactional capabilities. Another issue that we believe must also be addressed is that of access policies, related not only to the concepts manipulated by an application, but also to the application itself. In a single-party non-distributed world, most of the times one may assume that, as a developer, one is granted (or may be able to get) full access to all of the application code. In a multi-party distributed environment like the Web, one cannot assume such broad levels of access. Most of the times, a developer will be faced with the problem that he does not have access to analyze all the code that he is using, simply because some of the application code belongs to third-parties and, although they are eager for others to use it, they are not willing to share the knowledge underneath. So, the only alternative is to work with incomplete information at a declarative level (which is an issue in itself) or request assistance from others. To account for the latter, a development environment needs to be both distributed and collaborative. All this must be realized using a non-centralized (peer-to-peer) architecture. The complexity of such an (heterogenous, reactive and distributed) execution model, will render it useless, unless an adequate development environment exists (as demonstrated by the slow rate of adoption of results obtained in the field of ADBMS). In order to define such a generic model, one must take into consideration not only how it will work, but also how it will be used. The best way to ensure that such a development environment can be built (without having to redefine the execution model, or patch it up at a later stage) is to define the execution model side-by-side with the implementation of a prototype for an adequate development environment. That is what we propose to do: define a reactive execution model for evolution on the (Semantic) Web taking a WIDER perspective, thus fostering wider levels of adoption. Workplan *The present workplan was defined taking into consideration that all the results are to be completely integrated in the research in course under the European Project REWERSE (Reasoning on the Web with Rules and Semantics), in which the candidate participates (since January 1st, 2005, as coordination assistant of Working Group I5 - WG-I5). REWERSE is a research "Network of Excellence" (NoE) on "Reasoning on the Web" that is funded by the EU Commission and Switzerland, over 4 years, within the "6th Framework Programme" (FP6), Information Society Technologies (IST), Priority 2 under the project reference number 506779. REWERSE addresses the IST strategic action line "Semantic-based knowledge systems". REWERSE has started on March 1st, 2004. WG-I5 of REWERSE (coordinated by the candidate's supervisor) is dedicated to Evolution and Reactivity on the Semantic Web, focusing on issues like: investigate reactive behavior models that lead to evolution of data, rules and ontologies on the Web; investigate communication models to support these reactive models; investigate ways of dealing with inconsistency issues raised by global (and local) evolution in an heterogeneous environment like the Web; develop case-study scenarios for evolution based on reactive behavior; build prototypes for both showing how to manage and specify evolution, and to cater for reactivity on the Semantic Web. Since the beginning of 2005, we have been specifying (together with other REWERSE WG-I5 participants) the base model to provide reactive behavior in the Semantic Web [2]. The current results of this work have already been presented namely to the Semantic Web research community [13]. We are currently developing several prototypes in order to experiment and improve this model, and initial results have already been submitted and accepted in the specific field of Reactivity on the Web [6]. A first prototype (r3: Resourceful Reactive Rules), already at the ontology level, was completed by August 2006, and is available at http://rewerse.net/I5/r3/. Current results (summarized in [1]) point to the need for an extended generic model (including also a more formal definition). The work on this prototype also made it obvious that current development environments are quite limited when dealing with multi-party distributed reactive execution models (just following the inference path, in order to test and demonstrate the prototypes, is a true challenge). By March 2007, we expect to have a clear formalization of this extended model and start researching the extensions needed in order to build basic tools to visualize, edit and monitor heterogenous reactive rules. This work will be carried out in the following 6 months, between March and August 2007, while evaluating the prototype against realistic application scenarios. An extended prototype is expected during the first quarter of 2008, that coincides with the end of project REWERSE. The next 3 months will be dedicated to evaluating and putting into perspective the achieved results. By June 2008 we expect to have an extended version of the formal model in order to cope with Visualizing, Editing and Monitoring Reactive Rules. If deemed necessary, the next year up to August 2009, will be dedicated to extending the prototype with rule execution monitoring functionalities, while at the same time investigating collaboration issues raised by the multi-party development nature entailed by the Semantic Web reality. In the last quarter of 2009, the formal model should be extended to cover these collaboration matters. We believe that debugging reactive rules is an issue in itself, but, in order to provide a foundation as solid as possible for future research in the field, we would like to dedicate the first semester of 2010 to consider this matter, just before writing our PhD Thesis, which is expected to be delivered during the last quarter of 2010. References *
| |||||||||||||||||
|
| 1 Guest |
| Contact | Last Update: 2008−03−20 12:50, Last Visit: 2009−07−04 08:11 Optimized for: IE5+ 800x600, Developed by: Ricardo Amador ©2009−2008 Ricardo Amador | 2009−07−04 08:54 Administration |
