Date: Thu, 25 Feb 93 13:59:05 GMT From: Martin Kirk Subject: X/Open Position on LIS and TA Dave, Attached is a paper setting out the X/Open position on the LIS and TA issues. I hope it may be useful in formulating a UK position. Personally, I would still like to see WG15 drop the LIS requirement and would be highly supportive of the UK proposing such a change. I regret that I won't be at tomorrow's meeting. Regards, Martin X/OPEN POSITION ON LANGUAGE INDEPENDENT SPECIFICATIONS AND TEST METHODS INTRODUCTION X/Open believes that it is in the interests of the IT industry and its users that specifications which have been widely implemented and gained general acceptance in the marketplace should be made available as base documents for standardization. To that end X/Open has developed its standards policy, and has provided several of its specifications to IEEE committees; three of these are expected to become IEEE standards in 1993. Our standards policy includes points which are relevant to this discussion: * X/Open shall cooperate with formal standards bodies to bring standards-based Open Systems to the market in a timely and effective manner. It shall make its work available to standards bodies with such release of copyright as is required to permit material to be incorporated into formal standards. * X/Open, and its Technical Working Groups, shall observe the rules of the standards bodies with which they work and shall offer reciprocal liaison as required. In working with the IEEE, X/Open has therefore contributed significant resources to assist in the work of the relevant committees, in order to enable the process to be 'timely and effective'. This has included helping with the development of language independent versions of draft standards, and the development of test assertions. This position paper is based on what we have learned in this time. Our experience has been that less work is involved in developing test assertions than in creating a language independent specification. As a rough guide, the language independence work appears to be two to three times that required for the test assertions. TEST ASSERTIONS We feel that test assertions are a valuable addition to a specification, especially since they can potentially be used in an automated process for developing test suites. We are keen that test suites should be provided for standards, since they provide the only objective way of measuring, and ensuring, conformance. If conformance is to be tested, it is helpful if the test assertions are developed, once, by those who developed the standard. We believe that market forces will require that conformance be proved, and that those who need to provide this proof will be motivated to develop tests. These influences should be used to drive priorities for the development of tests, rather than standardization procedures. We suggest therefore that the lack of test assertions should not prevent a standard being approved. X/Open is working with Unix International, under contract with MITI, to research technology which will automate the generation of test suites from formally expressed specifications. LANGUAGE INDEPENDENT SPECIFICATIONS We feel that the work of developing language independent specifications is valuable, but we have several concerns: Timing If a draft standard is based on the specification of products implemented by one or more suppliers, there will be already in existence a proven interface. To turn a language- dependent interface into a language independent form does not of itself add any value, but does cause delay, perhaps of 6- 12 months. Value The second concern relates to the value of the finished standard. If a 'thin' binding has been developed, any user will need to employ two books in order to understand the standard; apart from the practical difficulty, the likelihood of misunderstanding and error must be increased. Resources Finally, we have doubts about the skills and resources necessary for producing language bindings. The committee which produced the Language Independent Specification will almost certainly not have the resources to produce the bindings; similarly, if a language committee may be faced with many language independent standards, and could have neither the skills nor the available effort to do the work. It is likely that there will be very patchy implementation of language bindings. In conclusion, we would not argue with the principle that standards should be represented in a language independent form. However, where a standard is being developed from an interface which is defined in a standard language, we believe that a language specific specification should be acceptable for standardization. At the same time there should be discussion of the need for, and resources to produce, a language independent form and the associated bindings.