On Saturday, August 16, 2003, at 10:48 AM, Jeremy Boynes wrote:
Here is the crux of my point. IMHO, the javadoc/specs should capture all
requirements. The RI is not to be used as the final definition
of the spec.
Am I wrong?
The TCK is the final arbiter of the spec.
In this case, there is an ambiguity in the spec - the RI does one thing
which we think is weird. Provided what we implement conforms to the spec and
passes the TCK then we do not need to duplicate the behaviour of the RI.
The final arbiter is the spec. If we believe that a test in the TCK is not required by the spec, I believe we can challenge the test to have it removed.
-dain
/************************* * Dain Sundstrom * Partner * Core Developers Network *************************/
