History

Greg Warnes developed the "RStatServer system": as a tool for rapidly creating and deploying statistical applications. RStatServer integrates the open-source statistical system R with the open source web application system Zope. This initial system has been used to create various statistical web applications. The RStatServer project has been successful in increasing the ability of statisticians at Pfizer to develop and deploy web applications, and is very popular with users. However, the current system is still not sufficiently simple to allow the majority of statistical programmers to easily develop statistical applications. Further, the computational modules developed for use with RStatServer are not accessible from other (non-Zope) user interface systems (Pipeline Pilot, Spotfire, etc.)

The OpenStatServer Project was conceived by Greg Warnes and designed and developed by J C Lawrence, and is the next generation of RStatServer development. OpenStatServer is intended to make it easier for computational programmers to deliver custom web applications. OpenStatServer provides a simple interface for computational programmers to generate computational modules and allow those modules to be easily used by custom web applications developed using a variety of user-interface systems (Pipeline Pilot, Spotfire, etc).

Initial goals

The initial goals of the OpenStatServer Project are:

• Support stand-alone computational modules implemented in any of a variety of languages (R, SAS, S-Plus, Matlab, arbitrary executables, ...)
• Provide a straightforward supported method for obtaining module information (call mechanism, UI design-information, and programming documentation) using industry-standard description formats (WSDL, XFORMS, XHTML).
• Provide both session oriented (data is retained between calls) and stateless (all necessary data is contained in the call itself) interactions.
• Provide both interactive (a.k.a. synchronous) and batch-oriented (a.k.a. asynchronous) interactions.
• Mimimize the amount of effort that is needed to provide or to use a simple computational module.
• Retain the current abilities of RSOAP.

Current state

Very early Alpha release is pending. Features of the Alpha:

• Synchronous and asynchronous RPCs? supported along with status queries and results retrieval for asynchronous RPCs?.
• Simple API/RPC definitions in the Kiva supported.
• Simple Python computation services supported
• Simple R computation services supported
• Multiple systems ("Earwigs") in the computation pool supported.
• Multiple OpenStatServer servers fronting the same pool of Earwigs supported.
• Modestly fault tolerant. Suitable only as a proofcase and further development.
• A source release is expected imminently (pending Pfizer Okay).