[
Lists Home |
Date Index |
Thread Index
]
Understood. The term 'reference implementation' gets
tossed about lightly and it has legal import in some
organizations.
That said, it does mean that the implementations the
W3C accepts don't indicate much more than interest
in the specification/recommendation. So one has to
look to the vendor if not the source. Stating that
a rec should not be accepted unless it is implementable
is a weak constraint. I am satisfied that in the
majority of cases given the kinds of people who do
the W3C recs, they will be 'implementable'. What
many want to know is if it is worth implementing,
is it worth having an implementation, and given an
implementation, how does one sure it is compliant
and conformant. Then, is it fit for purpose.
In short, a spec may be fit for purpose and a particular
implementation may not. How does one determine which
is the case? The term 'goodness' was used at one
time when talking about the Unicorn tests for SGML
systems. We had a bad time determining when an
SGML system was worth investing in and did testing
for a DoD contract of various systems to assess that,
scoring each against a series of tests that we as
SGML experts asserted were revealing.
Neither a reference or a sample implementation will tell
one that either.
len
-----Original Message-----
From: Michael Kay [mailto:michael.h.kay@ntlworld.com]
> Are the W3C implementations *reference* implementations
> or *sample* implementations? IOW, how is the implementation
> tied to the specification in terms of features and proof
> of conforming and compliant implementation? Are any public source?
> Are they tied normatively to the spec or informatively?
They are sample implementations, W3C doesn't endorse one implementation
by giving it a seal of approval as a reference implementation.
>
> Anyone can code anything. That doesn't mean it works as a
> goodness proof or can be used to test, validate, or verify
> conformance and compliance.
I think the W3C process is based on the assumption that a Rec should not
be finalised until it has been proved to be implementable, and the
existence of an implementation would seem to provide some proof that it
is indeed implementable.
Of course the fact that an implementation exists doesn't prove that it
is fit for purpose (which is what I assume you mean by "goodness"), but
it is evidence that someone, at least, considers it worth the cost of
implementing. And of course it's not a conformance test - that's a quite
separate process.
Michael Kay
|
