We are happy this week. Alain’s last paper, Personalised code generation from large schema sets for geospatial mobile applications, has just been published in the Springer Computing journal. It is the third and last article derived from the Alain’s PhD thesis which have been published in international journals. Let us recall the series of these important papers.
Researchers are problem-solvers. No doubt. But researchers are problem-seekers. No doubt too. Any research project or scientific work in general aim to shed some light on a (well-recognized) problem or issue. This is especially true in the context of a PhD thesis, where students struggle at early stages of their PHD period to find clear, convincing motivations in the form of identifiable issues which will then frame the scope and objectives of the research activities conducted over the rest of their doctoral period. Alain’s first key paper, which appeared in 2012 in IJGIS, formally identified a motivating problem in the context of developing geospatial applications for mobile environments: the great complexity of the data model (XML schemas) associated to some OGC services was a barrier to boost the creation of OGC-compliant apps. A developer should manually code XML parsers which were able to handle potentially tens of hundreds of XML entities to simply interpret an XML payload contained in a pair of HTTP request/response calls (e.g. a getCapabilities document). As it even happens with successfully standards like the HTTP protocol, some characteristics are hardly used in most HTTP implementations. The same holds for OGC services in general. We particularly discovered that only 30% out of all OGC Sensor Observation Service (SOS) functionality along with the associated XML entities of the data model were actually implemented in online SOS service instances. In consequence the remaining 70% were fully useless for XML parsers.
The proposed solution was to simplify the original service’s data model schema to the actual set of XML entities implemented by service instances. By doing so, a developer only needed to create a XML parser for handling a reduced and meaningful set of XML entities for a concrete service instance. Besides the simplification algorithm, we provided a sort of code generator that took the previous simplified data model schema and generated Android Java classes, i.e. a XML parser, to automate as much as possible the development of geospatial apps for a given OGC service instance. This 2-step methodology was the main contribution of the Alain’s thesis that’s described in the Computing’s paper.
The third article was a follow-up experiment to explore whether or not different data formats (e/g XML, JSON, compressed XML, etc.), different types of mobile devices, and communication networks altogether impacted on the effective transmission and performance of observational data between SOS service instances and mobile clients. Even though XML and JSON encoded messages varied in size, the main take-home message was in the way observational data was actually organized into a payload message. The Observations and Measurements (S&M) data model did not pay enough attention to the excessive repetition of bits of textual data for example. This got even worse as the volume of observational data increased up. Simple hacks, still relevant today, would reduce considerably unnecessary text overhead, as we suggested in the Sensor’s article.
- Posted by geoadmin
- On 9 April, 2014
- 0 Comments