minimum vs. reference implementation

In many areas, there are many ways to skin a [particular guideline] cat and
it is hard to choose among them as to one which should be called "minimum."
 Calling something a "minimum implemention" will lead people to infer that
_all_ implementations must do _at least_ that, i.e. they all must do it.

An alternate model is that offered by, for example, the "reference design"
of a cell phone done by the Trace Center as part of the Section 255 process
<http://trace.wisc.edu/docs/phones>.  Here the message is not "This is the
minimum you must do," but rather "See, it is indeed readily achievable
because you _could_ do it like this."  The Working Group would have to
convince itself that the example offered is indeed conforming.  But the
User Agent itself is not required to support this technique to be
conforming.  It can elect another approach which meets the performance
requirement stated in the guideline.

For guidelines where the language "minimum implementation" makes us
struggle, the working group should feel free to drop back and identify an
"example conforming implementation" which is felt to be readily achievable.

Certainly in the case of Guideline 2.1, the idea of a property sheet for
the current object is something that seems easy to do, and seems to satisfy
the guideline.  But I would find it hard to call it a "minimum
implementation" because I don't feel it is a univerally-required function
where the UA implements something better.

Al

Received on Saturday, 22 April 2000 14:23:20 UTC