- From: Maciej Stachowiak <mjs@apple.com>
- Date: Fri, 06 Feb 2009 17:31:37 -0800
- To: Ian Hickson <ian@hixie.ch>
- Cc: James Craig <jcraig@apple.com>, public-html@w3.org
On Feb 6, 2009, at 4:10 PM, Ian Hickson wrote: > > On Fri, 6 Feb 2009, James Craig wrote: >>>> >>>> 6. If no caption information can be determined from the previous >>>> steps, user agents MAY assume the image is a key part of the >>>> content >>>> lacking a textual equivalent in the document >>> >>> What does this require? Could you describe a test page that >>> demonstrates how to test whether a user agent is complying to this >>> requirement or not? >>> >>> There's no point adding requirements that aren't testable. >> >> Do you need to test an RFC 2119 "MAY"? > > It has to be testable, otherwise it's meaningless as a requirement. > MAY > requirements (and actually even SHOULD requirements) aren't likely > to be > tested in a conformance test suite, though, no. I don't think MAY-level requirements need to be testable, because they are not really requirements. Rather, they explicitly grant freedom to the relevant conformance class. For example, if you give an explicit algorithm in pseudo-code for some calculation, and that pseudo-code is known to be a slow but clear algorithm, you could say something like "User agents MAY calculate value X using other algorithms, so long as the answer is the same as that produced by this algorithm". That can't be tested, but it's not an invalid requirement, in my opinion. (Of course, permission to use faster but black-box equivalent algorithms generally goes without saying, but I can imagine similar cases where a MAY has no directly observable effect.) That being said, I am not sure the specific statement James suggested is necessary. Regards, Maciej
Received on Saturday, 7 February 2009 01:32:19 UTC