Re: [WebIDL] cycles in [PutForward] chains

On 8/4/12 1:33 PM, Tab Atkins Jr. wrote:
> Since the audience for that requirement is spec authors themselves, I
> think the presence of a bare MUST NOT is sufficient.  If a spec author
> violates it, someone can point it out to them, and they can change the
> spec.

Indeed.  In Gecko's WebIDL bindings, we're simply treating every MUST or 
MUST NOT violation as a fatal error in the IDL and refusing to compile 
it, period.

I would personally be fine with making that a general WebIDL 
implementation requirement: that an implementation of an interface that 
violates the WebIDL specification simply not be possible; if one exists, 
it's an implementation bug (in addition to a bug in the relevant interface).

-Boris

Received on Saturday, 4 August 2012 20:00:44 UTC