|Were we at a recording of 'This is Your Life', and not the ASC's September conference on 'Open Standards'? As each speaker rose in turn to hail the influence of triple-s on their lives, only the famous red book was missing.
Peter Wills (Mercator) opened as the bemused but proud father, as the one that floated the idea for an open standard for survey data interchange at the MRS Conference ten years ago. Geoff Wright (Computable Functions), one of the three founders of the group that developed triple-s, was that special teacher who coaxed and stretched the standard to be what it is today. "We deliberately went for simplicity all the way through," he explained. "You need low barriers to adoption otherwise it is a waste of time." Stating that today, some 25 developers have adopted it, he noted that the group is now "spending more of our time supporting it than developing it." Despite that, version 1.2, supporting filters and multiple languages, was published this year and a version 2 is being considered.
Keith Hughes (Merlinco), life-long drinking buddy to triple-s, discussed the close links between triple-s and OpenSurvey's two XML standards initiatives: AskML, for complete survey instruments and TabsML for aggregated survey data. He explained the dangers of leaving the thinking to individual vendors:
|"Many manufacturers are producing XML definitions - FIRM, Perseus, SPSS MR and, soon, Blaise - but they all differ widely." For Hughes, the challenge of harmonising logic and routing in AskML is the greatest. Appealing for help, he said. "It is a very big task. Future progress depends on greatly expanding the project or possibly transferring to an institute such as a university."
Ed Ross, who in his former Quantime days, famously stated his products were the standard, cameoed as old sparring partner turned loyal buddy. In a fascinating paper, Ross scrutinised the concept of data models with reference to the three current open interface initiatives from Blaise, SPSS MR and triple-s.
Praising SPSS MR's Data Model as "Breathtakingly ambitious and hugely flexible", he expressed concerns that "it can act as a black hole. Having no representation of logic is a real limitation and the complexity of the model is causing problems of understanding as well as some performance issues." While Blaise was simple, "it is hard to change the metadata, hard to
|uncouple from Blaise [proprietary] scripting and is somewhat clumsy." Hierarchical data, Ross noted, is still ignored by all three. He also put his finger on the greatest barrier to open standards, when observing that the uptake of triple-s by the software majors was often "inversely proportional to the size of the company."
A trio of case studies rounded off the day. Birju Jani of CfMC related this US manufacturer's experiences when implementing triple-s, Sayed Zia presented a triple-s inspired XML metadata project from Seattle's University of Washington, and BMRB's Mark Pietronave reported the details of how triple-s regularly saves months of work on its massive TGI project in China. It seems persistence and enlightened self-help is paying off as triple-s, at last, gains important international recognition.
Tim Macer's website is at www.meaning.uk.com