On 10/6/11 12:11 PM, Adam Barth wrote:
It sounds like you're arguing that it's better for developers if we
fail fast and hard
In some cases, yes. It's a tradeoff in every case, obviously.
A meta-issue: if you disagree with the spec text when implementing
something, silently implementing something else seems strictly worse
than raising a spec issue (and still implementing something else if
desired).
Especially for things that you're planning to implement unprefixed.
Likewise for cases when the spec is unclear, etc. What's the point of
having implementations early in the specification process if they don't
actually provide feedback and instead only serve to lock in behaviors?
-Boris
