The value of Candidate Recommendation (10:40 -0700)
I'd like to share a point I made yesterday the CSS working group meeting, in some side conversations about dropping some features [editor's draft, not yet official] from css3-color that some people have heard were implemented somewhere (without any evidence). It didn't generate any controversy (yet), but I though it's worth mentioning.
The W3C process generally requires two implementations of every feature to advance a specification to Proposed Recommendation. The best-known reason for this is showing that the spec can be implemented (along with the other specs that it must interact with). But another reason for this (the one that I brought up) is that the feedback from the implementors improves the spec. Many implementors, when implementing a new specification, tell the working group where the spec is unclear, where it contradicts itself, where it should have defined things more clearly, etc. This feedback improves the specification and helps ensure that other implementations will interoperate. If a spec has implementations written by implementors who did not provide this feedback, I'm hesitant to allow those implementations to count for the purposes of entering Proposed Recommendation, since, although the spec has been implemented, it hasn't been improved (by the feedback from implementors) to the level of quality that a Proposed Recommendation should meet.
The concrete result of this point in yesterday's discussion was my suggestion was that even if these features have been implemented, we should still drop them from this version of the specification.
