RE: [HTMLWG] CR Exit Criteria redux

On Thursday, September 20, 2012 9:23 AM, Maciej Stachowiak wrote:
> On Sep 20, 2012, at 9:11 AM, Adrian Bateman <adrianba@microsoft.com> wrote:
> > Since the goal of testing for CR is to confirm that it is possible to
> > independently implement the spec and achieve interoperability, why is one
> > month of "stability"
> > necessary? If someone ships an implementation that passes tests and
> > demonstrates that interoperability has been achieved, why is it necessary
> > to wait another month?
> 
> That clause is copied from CSS WG's stock conformance criteria, which were
> used as a model. I don't know the original motivation. I suspect it is there
> to limit evasion of the "not experimental" clause. It's also my experience
> that code that someone wrote yesterday is in general less credible evidence
> that something is implementable than code that has been lived on for a while.
> Just-written code is likely to have showstoppers. Note that the requirement
> is that the feature has been *implemented* for a month rather than *publicly
> available* for a month. So it's likely not necessary to wait a month after
> shipping in most cases.

Thanks Maciej. I personally don't have a problem with "experimental" implementations.
We're testing the spec not the implementations, after all, and it's hard to know
for some projects whether a nightly build is experimental or not, but I think I
understand the motivation.

I'm always in favour of not stating things that don't affect the outcome. Does
"implemented for one month" mean no bugs have been found for a month, or that it's
a month since I started working on the feature, or something else? And if I can
just say it was a month ago it doesn't really impose much of a bar and certainly
hard to evaluate as a MUST. I don't think this is a useful addition.

Received on Thursday, 20 September 2012 16:39:27 UTC