From kh@athens.cs.waikato.ac.nz Wed Aug 10 21:09:34 1994 Received: from grace.waikato.ac.nz by dkuug.dk with SMTP id AA24288 (5.65c8/IDA-1.4.4j for ); Tue, 9 Aug 1994 23:06:39 +0200 Message-Id: <199408092106.AA24288@dkuug.dk> Received: from athens.cs.waikato.ac.nz by waikato.ac.nz; Wed, 10 Aug 94 09:06 +1200 Received: by athens.cs.waikato.ac.nz (16.6/16.2) id AA13506; Wed, 10 Aug 94 09:09:34 +1200 Date: Wed, 10 Aug 94 09:09:34 +1200 From: Keith Hopper Subject: What is conformance? To: sc22wg15@dkuug.dk X-Charset: ASCII X-Char-Esc: 29 Hi, Mmmmmm! I detect a certain amount of dissatisfaction in message 386 about what is perceived to be the "Modula-2 view of conformance". Having a hand in both camps I feel that there has to be a misunderstanding somewhere, since I have a single view -- which fits both the POSIX version being developed and the M-2 one in the DIS. I believe that conformance is about the ability to assign a meaning to a program. In terms of a programming language standard the conformance of an implementation either allows this or it doesn't. We have all seen meaningless programs, I'm sure. In terms of a binding conformance says that an implementation must bridge the gap (small or large) between the semantics of the package to which a binding is being made and the semantics of meaningful code written in the language X. It is a Janus-like conformance requirement. For a package, the matter of conformance is independent of the language and is therefore more of an open contract between package provider and user --- nevertheless only ascribing meaning to some action if the user meets any pre-conditions -- his part of the contract. This view of conformance seems to be consistent in that at the language level there is either a meaningful chunk of source text -- or not, a binding either provides transparent mapping of meaning -- or not, a package either satisfies itd 'crontactual' obligations -- or not. Testing for these is not something which can be done independently -- there has to be some form of integration testing, particularly for bindings. Doesn't this seem sense? Regards, Keith