Objectives for the Web Service Hackathon

Background

Databases in lifescience area have enormous amount of data, still growing and becoming more multi-faceted. To provide integrated services , the way just centralising these databases, like by mirroring each databases, is insufficient but also the way using directly over distributed environments is strongly needed to be constructed at really serviceable state.

This situation brought on SOAP/WSDL web services to be launched at institutes around bioinformatics like NCBI and EBI abroad, and DDBJ XML, KEGG API, PDBj, CBRI in Japan, providing respective services handling many tools and databases.

Also, by myGrid project and BioMOBY project, efforts has already been made to provide foundation to use those web services in an unified way, which brings situation very practical to implement integration by web services technology around this field.

Main problems are, the variety of each specification and naming convention, and unstandardised data structure to be passed. Additionally, handling for cases like temporal service down or whatever error occured while executing a job just rely on each provided servers which tends to leave insufficient specification documentation.

Existence of different usages among each services, or situation compelling user to exchange data types and to handle exceptions for their own is really inefficient. Furthermore, the number of serviced tools and databases providing web services are still few, which makes it quite severe to achieve constructions of workflows through these services for the moment.

Efforts in DBCLS

Thus we at DBCLS have planned to search usage of existing web services and of those data structures to construct proxy-like server, which aim to provide unifed and consistent naming conventions and usages. To make this, we would first consider;

• Services be well documented.
• Owe error handling at server side as much as possible.
• Servers to be accessible by languages widely used like Perl, Ruby, Python, Java and more as many as possible.
• Owe data exchange by server side to make communicate between many servers servicing web services.
• Pipelines over several steps be done on server side to be effective enough.

By conducting our proposal, not just providing environments exeeded in usability to many researchers, it can act as an infrastructure for constructing workflows which accordingly provide each centers a dramatical increase of internet access counts.

At the same time, consideration is needed over what categories of services to provide, those quantity and quality. Therefore it is desirable to parallelize working on providing service along with developments of integrated databases and tools at DBCLS.

Objectives of this Hackathon

To lead to a standardisation of;

• naming convention of methods
• data structure passed among services
• job controlling

which at present differ within each providers.

Therefore, we are going to held developer's meeting in January or February of 2008. There, core developers and key members at home and abroad related to each web service providers, including BioMOBY, Open Bio*s (ie. like BioPerl?), would be offered to attend this meeting, staying for about a week in Japan.

So far, data type (class design) for bioinformatics has been defined for each project of Open-bio (like BioPerl?, BioPython?, BioJava?, BioRuby) respectively. Though to benefit interoperability, it seems to be nice to define standard specification of objects based on web services' class, to comply with it.