Vol.3 No.1 2010

75/110

Research paper : A field-scientific approach to Clinico-Informatics (Y. Kinoshita et al.)−72−Synthesiology - English edition Vol.3 No.1 (2010) from, for instance, the market.e) [Blind test] Take an old development with some properties (nice or badit can be a fault). It is assumed that the technology of concern can ease the findings of those properties. The properties are hidden from the research team who tries to clarify them using the technology of concern and evaluating the results of the process. For instance, take an old development with some known, recorded faults. These faults are hidden from the research team who tries to find them using model checking. This can be a good way to evaluate the value of model checking.Who exercises the technology of concern? There are at least three possibilities.i) By a team composed only of members of the research teamii) By a team composed of a mixture of members of the researchers and those of the engineersiii) By a team composed only of engineers of the industrial partnerIn early stages, one tends to follow i) to display the effect of the technology of concern; then the technology is gradually transferred to the industrial partner by means of ii), and finally, the trial experiment is conducted by iii) to evaluate the cost.6 Two case studiesAmongst many fieldworks of ours, there were two cases that were continued for more than three years which were concerned, as a result, not only with a short term goal but also with a middle term one. In this chapter, these two works are presented and we try to evaluate their outcome.6.1 Industrial partner PThe fieldwork done jointly with our industrial partner P started in response to a demand by P for introducing model checking into their development process. P reached model checking after their search for a method of developing reliable software with high quality.1. A “Model experiment” was conducted for a small piece of software of P. Model checking was conducted for about one month, and an engineer on the side of P learned the process of verification using model checking, while the research team learned how to read the specifications written in P and the basic domain knowledge. This was repeated several times.2. A “Blind test” was conducted after repetition of 1. All the faults that should be discovered were found by applying model checking. A good set of examples showing the effectiveness of model checking was obtained through this Blind test.3. Up to this point, the work of model checking itself was done by the research team. At this point, the project decided to write a manual which enables engineers of P to conduct model checking without assistance of experts. To that end, model checking is now done jointly by the research team and engineers of P and this joint team repeated “pre-development experiments” and “post-development experiments,” several times to write down a guideline for a verification process using model checking, and a manual to verify a module in nine days. The guideline was written jointly but the manual was completed solely by the engineer team.Unfortunately, we have not received any report from P about how model checking was introduced into their own development process after this fieldwork was completed. We assume it has not been deployed in a large scale, to our disappointment. Note that, however, P is a world wide enterprise having more than a hundred thousand employees, so for them to deploy a new technology would itself be a huge project.For a technology transfer project to succeed, there are various points that must be considered, other than technical problems that would be solved by trial experiment as discussed above. We realized through this fieldwork that the research team must set up a view about intellectual properties; that is, it must have a fixed strategy concerning which results would contribute to intellectual properties and which results would contribute to academic publications, and such strategy should be set up at an early stage of the whole project. Take, for example, the writing of a manual as described in 3 above. The manual was written only by the engineers of P and no one from the research team participated in the writing. This, however, prevented information sharing between the engineers and researchers. In retrospect, the researchers should have supported the manual writing much more because some information which must be written down in a manual is not so important as an intellectual property or as a result of academic research but, at the end of the project, we tended to regard every information written in the manual as valuable either as an intellectual property or an academic result.6.2 Industrial partner QA fieldwork for technology transfer jointly with our industrial partner Q started when engineers of Q became interested in model checking, and the top management of Q dispatched one engineer to AIST for two years.While the fieldwork with P was done with the intention to write a manual to propagate model checking to exclude the dependency on experts, in the case of Q, technology transfer

※このページを正しく表示するにはFlashPlayer9以上が必要です