- From: Alex Rousskov <rousskov@measurement-factory.com>
- Date: Mon, 6 May 2002 15:28:17 -0600 (MDT)
- To: Charles McCathieNevile <charles@w3.org>
- cc: spec-prod@w3.org, w3c-query-editors@w3.org, www-qa@w3.org
On Mon, 6 May 2002, Charles McCathieNevile wrote: > ... not being able to test whether a specification is being met > means that it is less a specification than a general description > of an idea. I do not think the above is true in general. For example, there are numerous working implementations of HTTP specs while many HTTP statements cannot be tested in a pragmatic way. The primary goal of a specification is to enable building [compliant] implementations. This goal is different from enabling [compliance] tests. It would be great if all specs were 100% testable, but I do not think it is possible in practice, regardless of the specs language. My belief is based on a simple fact that both black- and white-box testing techniques cannot achieve 100% coverage of a complex program implementing the specs. Testability should definitely be a priority, but it would be sad if we get fewer good specs by accepting rigid and expensive testability requirements. A poor solution is often worse than a simple acknowledgment of the problem. Alex.
Received on Monday, 6 May 2002 17:28:22 UTC