- From: Sean Owen <srowen@google.com>
- Date: Fri, 8 Feb 2008 12:13:06 -0500
- To: "Paul Walsh" <paul@segala.com>
- Cc: achuter.technosite@yahoo.com, "MWI BPWG Public" <public-bpwg@w3.org>
On Feb 8, 2008 11:42 AM, Paul Walsh <paul@segala.com> wrote: > I agree. It was originally agreed that the mobileOK basic would be > made up of tests which were relatively easy to comply with. Whether > these were machine-testable or not, was not relevant. However, > further down the line, this changed for some very strange reasons. We > now have mobileOK basic which is machine-testable - even if they > aren't necessarily easy to comply with. Again the logic was: it's overwhelming useful to provide something machine-testable. So let's put into Basic everything that is easy to comply with or important, and also machine testable. But then there were repeated claims that the suite was too easy -- this is the reason you are looking for. So it became everything machine testable. I think that's fine. It's a simple condition, and has resulted in a useful specification and tool. > "since they would need a certified mobileOK evaluator to test" > > What do you mean by this? Well let's say we put out a suite of tests. Does X pass it or not? if it's machine testable, then you run some presumably correct implementation of the machine tests on X. You could implement it yourself but you don't have the time, and are anyways not sure you will come out with the "right" answers. If anyone then cares to ask why you pass, it's not so great to say, well, I wrote code that says I pass. It would be much more credible to say that a third-party system passed you. Analogously, for a human testable suite, you might suggest that individual developers are going to effectively implement mobileOK Pro themselves by becoming an expert on mobileOK Pro, executing the tests and assuming they're right, and then feel it's useful to say, I pass the tests because I say so. I assume this model really also entails some third-party "implementation", and that involves humans, and involves time and money. >From there we get into the same old circles... there is no doubt some market for this, but at the moment, given that mobileOK Basic is not exactly taking the world by storm, I think we do need to honestly ask who mobileOK Pro is for. Is there at least one potential customer someone can identify? I do hear strong convictions here that this is not only important to do, but that it's somehow dangerous not to do it. I would just like to continue to understand why.
Received on Friday, 8 February 2008 17:13:19 UTC