1. Model-Driven Web Feature Service A Way Towards Enhanced Semantic Interoperability Peter Staub, ETH Zurich FOSS4G 2007 – Victoria B.C., September 26, 2007

2. Peter Staub, ETH Zurich2 Overview Data modeling and semantic interoperability: Fundamentals Research project mdWFS Why a model-driven WFS? Enhancements in semantic interoperability Concepts, methods Prototype implementation

3. Peter Staub, ETH Zurich What is Interoperability? 3 Swiss plug does not fit American socket So, an appropriate adaptor plug (and a voltage transformer) is needed! I – Modeling and Interoperability Fundamentals Interoperability: Format Support + Data Model Transformation 230V ~50  110V ~60 

4. Peter Staub, ETH Zurich Data Modeling and Semantic Interoperability 4  PIM>PSM Mapping GML App. Schema Logical/Physical Schema A (PSM) > Logical/Physical Schema B (PSM) >  Encoding GML Instance Document Database A, Transfer format A > Database B, Transfer format B > MOF > Metamodel CSL X > Metamodel CSL Y > Schema Translation > Conceptual Schema Mapping > :Instance Translation > Conceptual Schema A (PIM) > Conceptual Schema B (PIM) >  Formal mapping Real World Application Domain „Model“ B Application Domain „Model“ A I – Modeling and Interoperability Fundamentals

5. Peter Staub, ETH Zurich Project mdWFS Web-based cross-border GIS: a regional planning use case What are the shortcomings of existing approaches? Overcoming these shortcomings with a model-driven WFS 5 „Model-Driven Approach for Accessing Distributed Spatial Information Using Web Services“ II – Project mdWFS

6. Peter Staub, ETH Zurich Problem and Requirements Shortcomings of OGC Web Services [OWS]: OWS allow for syntactic interoperability but not for semantic interoperability Preconditions for a web-based semantic transformation on the conceptual level Machineable CSL for data models Formal language for expressing schema mapping rules Web service interface Web service requirements Access to geospatial data based on conceptual source data schema and conceptual target data schemas Interoperability with existing OWS: WFS 6 II – Project mdWFS

7. Peter Staub, ETH Zurich mdWFS Interface Idea: model driven Web Feature Service mdWFS Extension of OGC WFS specification Ability to store and deliver conceptual models (UML/XMI) New operation DoTransform() enables WFS carrying out semantic transformations Each semantic transformation results in an according set of WFS feature types WFS is configured by target schema at conceptual level! 7 II – Project mdWFS

8. Peter Staub, ETH Zurich WFS Protocol Extensions for mdWFS Specification of communication protocol for mdWFS  new request parameter SERVICE=mdWFS (default: WFS ) Specification of WFS protocol extensions GetCapabilities WFS: FeatureTypeList mdWFS: SchemaList DescribeFeatureType WFS: XMLSchema (transfer format schema) mdWFS: XMI (data model) GetFeature  new request DoTransform performs semantic transformation and configures a standard WFS 8 III – mdWFS: Concept and Methods

9. Peter Staub, ETH Zurich What is needed? Metamodel: UML 2 profile for geospatial applications (e.g. Interlis CSL) Model Parser to convert models (textual notation) into XMI 2.1 Language specification for semantic model mappings [UMLT]: Extension of UML, Activity diagram Model editor to create conceptual model mappings using UMLT Transformation operations to execute semantic transformations 9 III – mdWFS: Concept and Methods

10. Peter Staub, ETH Zurich Schema Mapping Language UMLT Requirements: Comprehensible for non-computer scientists Metamodel, HUTN (Human Useable Textual Notation) Visual, textual AND XML (i.e. XMI) representation Application of international standards Basic principle: Independent extension of UML metamodel Metamodel of UMLT as UML 2 model EBNF (extended Backus-Naur form) for UMLT grammar specification 10 IV – UMLT Specification

11. Peter Staub, ETH Zurich Schema Mapping Language UMLT: Superstructure 11 IV – UMLT Specification

12. Peter Staub, ETH Zurich Schema Mapping Language UMLT: Elements 12 IV – UMLT Specification

13. Peter Staub, ETH Zurich UMLT Example: Source and Target Data Model 13 V – Prototype Implementation

14. Peter Staub, ETH Zurich UMLT Example: Transformation Model 14 Centroids V – Prototype Implementation

15. Peter Staub, ETH Zurich UMLT Example: Trafo. Model in Model Editor 15 V – Prototype Implementation

16. Peter Staub, ETH Zurich 7 ili2ora Prototype Implementation 16 Client = Target System Model B: UML/XMI ORACLE Server = Source System Model A: UML/XMI 0 mdWFS WFS 1 Request model catalogue 2 Provide model catalogue 3 Order required model 4 Send ordered model A [ XMI ] 5 Establish Model Mapping A  B ili2ora SQL/Java 8 ABAB 6 DoTransform -Request Model B + Trafo. Model A  B [ XMI ] deegree 9 Model B DoTransform -Response 10 GetCapabilities DescribeFeatureType GetFeature 11 V – Prototype Implementation

17. Peter Staub, ETH Zurich Conclusion, Lookout 1.Current evolution of GIS: use of CSL (precond. for semantic Trafo) But: metamodels/profiles and „modeling styles“ differ in detail 2.mdWFS: semantic transformation is established at conceptual level; independent from any format or (DB-)system 3.Concept based on GI standards allows for integration into existing (OWS based) infrastructures To do: Finish implementation of transformation operations Further development of UMLT Prototype evaluation in the context of geodata infrastructures 17

18. Peter Staub, ETH Zurich Thank you very much! peterstaub@ethz.ch Acknowledgements German Federal Agency for Cartography & Geodesy Swiss Federal Office of Topography TU Munic project team Colleagues at ETH Zurich 18