Re: Feature List

Charles McCathieNevile wrote:
> 
> On Wed, 17 Dec 2008 12:15:36 +0100, Phil Archer <phil@philarcher.org> 
> wrote:
> 
>> Anthony Kukurikos wrote:
>>> ...One question: should we have tests for the
>>> cardinalities of each include/exclude element? It is not a matter of
>>> workload as it is trivial, I just don't know if it facilitates the
>>> readability of the TS doc (which is of course an important matter but 
>>> not as important as its usefulness).
>>
>> It's a matter of striking a balance between 'proving everything works' 
>> and going over the top with a separate test for every last thing. I 
>> lumped the IRI constraints together as a compromise on this. It's only 
>> in/excludepathcontains and in/excluderegex that's allowed more than 
>> once anyway - the rest are all 0 or 1.
>>
>> If you have a good test to hand, good - use it, but let's not over do it!
> 
> Hmm. I think it is good to use any tests we have - they all help improve 
> the quality of implementations (or find bugs). The balance question is 
> more one of judging whether we even have enough tests of the different 
> aspects to make a reasonable claim that we are done...

No argument there Charles. All I'm getting at really is that, for 
example, the first 'feature' is "Basic structure of a POWDER Document" - 
well, that covers a bunch of MUSTs and SHOULDs immediately following 
example 2-1 - which the validator tests for. I'm hoping we can avoid a 
feature list that has a line for every RFC2119 keyword ;-)

Working on thus list today, I admit I've fixed a few bugs in the 
validator and processor - so the exercise is doing its job!

Phil


-- 
Phil Archer
w. http://philarcher.org/

Received on Wednesday, 17 December 2008 15:41:17 UTC