Re: Normative status of author-only view of the HTML5 specification

On Tue, Jun 7, 2011 at 3:41 PM, Paul Cotton <Paul.Cotton@microsoft.com> wrote:
> Our goal is have the two documents agree - the author-only view is generated directly from the HTML5 spec.

The point of specifying some material as normative and some as
informative is so that you can have everything specified normatively
in exactly one place, so there's no *possibility* for contradiction.
Otherwise, why not make all of the informative parts of HTML5
normative instead?

On Tue, Jun 7, 2011 at 4:56 PM, Roy T. Fielding <fielding@gbiv.com> wrote:
> The audience of HTML5 authors and authoring tool makers is several orders of magnitude larger than the audience of browser makers. Given the way that the full spec is written, it will not be considered authoritative regardless of what is in the front matter.  If the author-only view is authoritative, then there will be pressure to both keep it in sync with the larger spec and keep the work at W3C.  If the author-only view is not authoritative, then there is no need for it to be in sync with a larger spec that nobody implements -- it will be replaced by documents that accurately describe HTML as implemented.

The author-only view is generated from the same source file as the
main spec, so no additional pressure is needed to keep it in sync with
the larger spec.

On Tue, Jun 7, 2011 at 5:25 PM, Roy T. Fielding <fielding@gbiv.com> wrote:
> The answer is that it is desirable for both documents to be authoritative or for *only* the author view to be authoritative.

Do you mean the author view should be authoritative for authors, while
the main spec should be authoritative for all other conformance
classes?  In particular, conformance checkers are a different
conformance class, and the spec contains important requirements for
conformance checkers that aren't in the author-only view and shouldn't
be (e.g., the HTML parsing algorithm).  So what you're proposing is we
have one document define the conformance requirements for authors, and
another document define them for validators?  That would be . . .
peculiar.

Received on Tuesday, 7 June 2011 23:02:43 UTC