Re: Correctly mapping role="search"

We're in sync. 

I just get uneasy of terms like mandatory because they tend to
lock things down before they are ready. As long as you meant it
in the sense of a working draft, yes, I am supportive of the
idea. 

The thing to avoid is someone taking the "must " in the working
draft out of context and insisting that that's how the world
should be forever going forward.

And I definitely prefer a concrete algorithm to a pile of
wishy/washy text, which is why I opened the message with
"algorithm = good"

Benjamin Hawkes-Lewis writes:
 > T.V Raman wrote:
 > > Algorithm == good --- mandatory --- at this early stage == asking
 > > for disaster.
 > 
 > Hmm.
 > 
 > Perhaps by "mandatory", you think I mean a formal Recommendation.
 > 
 > When I say "mandatory", I mean the draft specification should use 
 > language like "MUST" when talking about the algorithm, not talk in 
 > wishy-washy terms like "MAY" and "SHOULD". Such language would be 
 > relative to ARIA's status as a Working Draft. In so far as 
 > implementations could meaningfully be said to conform, they would 
 > conform to the draft as of a stated date.
 > 
 > If this is what you thought I meant by "mandatory", can you elaborate on 
 > how this would be "asking for disaster"?
 > 
 > If an algorithm is mandatory (in this sense) and implementors find a 
 > serious problem with it, then there's time to change it long before ARIA 
 > becomes a Recommendation. On the other hand, if there's no mandatory 
 > algorithm:
 > 
 > * It's hard for implementors to know how to apply the roles, since what 
 > they mean is merely suggested and not defined. role="main" might be the 
 > main area for the ancestor role="document", or it might not.
 > 
 > * It's hard for implementors to give feedback, since it's not clear what 
 > the spec is saying.
 > 
 > * We have lots of people encouraging very early adoption of ARIA. 
 > Without mandatory algorithms, implementations by early adopters might 
 > differ substantially, potentially preventing a reliable algorithm being 
 > defined later. Others may disagree, but I'd suggest even a suboptimal 
 > but well defined and widely implemented algorithm is preferable to a 
 > perfectly designed conformance requirement produced late in the day that 
 > nobody can implement without breaking existing content.
 > 
 > --
 > Benjamin Hawkes-Lewis

-- 
Best Regards,
--raman

Title:  Research Scientist      
Email:  raman@google.com
WWW:    http://emacspeak.sf.net/raman/
Google: tv+raman 
GTalk:  raman@google.com, tv.raman.tv@gmail.com
PGP:    http://emacspeak.sf.net/raman/raman-almaden.asc

Received on Wednesday, 8 October 2008 16:57:31 UTC