The following issues need to be looked at as output from the review process:
- SOA4ALL plan to submit WSMO-Lite as a member submission to W3C as an extension of the SAWSDL recommendation. This submission should be a joint submission by SOA4ALL and LarKC
- The L2 language is currently in the pre-standardization phase in the CMS working group. It should be planned that L2 moves towards W3C as a member submission in the future (with the potential of making it a recommendation)
- We need to revisit how the L2 language is integrated within the project, we claim that it is the minimal representation language for LarKC but to date we have done no checking to ensure that at least the plug-ins developed within the project can minimally support L2
- Non-functional properties and preferences have been extensively researched in the Semantic Web Service community. It would be a waste of resources for LarKC to reexplore this issue without considering existing solutions. The work in the STI International CMS working group should be considered as a starting point
- SOA4ALL uses OWLIM as an underlying storage for triples as does LarKC, however they use a triple space implementation on top of OWLIM as the main integration point. Trying to ensure that the data layer in LarKC is someone in line with the data layer in SOA4ALL would probably make it easier for Ontotext to support the two projects in a clearer way and enable both projects to benefit from developments in the other.
- The OASIS SEE-TC is in the process of standardizing a reference architecture for a Semantic Execution Environment. Such a SEE is essentially used for findings, selecting, composing and invocation services and could be used as a LarKC decider for finding, selecting, composing and invoking LarKC plug-ins with SWS descriptions. More research into this should be done, with emphasis on the software components produced by the SOA4ALL project for finding, selecting, composing and invoking services described with WSMO-Lite.
- VOID is a vocabulary for describing the properties of an RDF dataset, such that crawlers and systems that want to use the dataset can understand its size and scope prior to retrieving it. Minimally all datasets produced by LarKC should be described in the VOID vocabulary. We could also consider allowing plugins to use parts of the VOID vocabulary to described the properties of the sorts of datasets they can operate over.
- Look at the alignment with the Semantic Vectors. Angus Roberts has already sent around a number of links to papers, namely:
Overview paper: http://www.puttypeg.com/papers/semanticvectors.v2.pdf
Various papers: http://www.puttypeg.com/papers/
- The SOA4ALL Studio provides a means for creating mashups of different services described with WSMO-Lite. We should explore using this tool as a means for creating scripted (or semi-scripted) LarKC workflows.