[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: Re: [sca-assembly] Concrete Exit Criteria for the SCA Assembly TC- Proposal
From the charter: http://www.oasis-open.org/committees/sca-assembly/charter.php#item-4 "The TC shall define concrete exit criteria that include at least two independent offerings that implement and are compliant with the all normative portions of specifications and demonstrate interoperability and portability as appropriate. Note that these are minimums and that the TC is free to set more stringent criteria. " The ambiguity of English gets in the way here a little. Does this mean:
What this discussion turns on, then, I think is the meaning of "implement", and "compliant". Is it sufficient for a vendor to declare that they've implemented and are in compliance? Or do we have to "prove" that in some way? I'm satisfied if a vendor that claims to implement a normative statement. As for "and demonstrate interoperability and portability", based on what I understand of the existing test suites, that two implementations that pass the test suite is a pretty darn good demonstration, and I wouldn't go further than that. I'm with Jim in that at some point we have to rely on the court of public opinion. From a pure "quality of work" perspective, I think if two implementations pass the test suite, and they comment about the normative statements not covered by the test suite, then we've acheived sufficient "code coverage" if you will, of the spec itself. I've expressed an ongoing concern that we've over-specified parts of SCA. Certainly, within the bindings TC, some of you have heard me express that point. To pick on a random point (really, I just looked at the normative statements and this one caught my eye - hadn't thought about it before), the "sca-contribution-generated.xml" file declared for deployment - how is this file an essential part of testing conformance? Since there's no language that I'm aware of that *excludes* other files from being processed as part of the contribution, why couldn't this detail be left to an implementation? Do we *require* this for the specification? If a side-effect of the "two implementations" constraint leads us to de-normalize (de-normativize?) some of the current required language, that might be a good thing. I'll reiterate what I said in the call. We should check with the implementations that we're aware of, and find out which normative statements they don't believe they've implemented. With that information in hand, we'll better understand our options for this question. (And I'm not volunteering to take on this task because I don't know who to talk to, and I'm not familiar with the existing test suite.) -Eric. On 2/22/11 12:08 PM, Jim Marino wrote: 74664AB4-833B-46BF-BFBC-56D0631456C9@gmail.com" type="cite"> |
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]