Re: ReSpec WebIDL authoring

Den 23. april 2015 08:52, skrev Dominique Hazael-Massieux:
> Hi,
> The way we define WebIDL in ReSpec (using <dl><dt><dd>, now qualified as
> "oldschool") is going to be deprecated down the line:
> The new approach (called contiguous WebIDL) uses proper WebIDL fragments
> with accompanying conventions for interlinking, cf:
> At the moment, the two modes are maintained; but clearly there are plans
> for a time where the oldschool mode won't be specifically maintained any
> more.
> Switching from oldschool to contiguous is a non-trivial effort (see for
> instance the patch that was required to the bluetooth spec [1]); I think
> there are likely ways to automate or semi-automate it, in which I could
> look if we determine we should switch.
> But the main question at this point is whether we should switch, and for
> which specs?
> Dom
> 1.

This seems like a wonderful thing to do - writing WebIDL instead of
trying to write something that respec maps back into WebIDL eliminates a
large class of errors, and should be easier to maintain in the long run.

But I don't want to be the first guinea pig....

I think the reference gives some of the answer:

The following aspects of WebIDL are not currently supported and are not
yet tested for:

* Static, inherit, stringifier attributes. <--- We use stringifier
* Static operations
* IDL wrapping (what's this?)
* Special operations (what's this?)
* Union types
* partial dictionaries  <--- We use partial dictionaries, I think.

Support for these features will likely be added later.


- We won't switch getusermedia (unless someone volunteers for the work)
- We'll switch webrtc when all the features we need are supported, and
someone else assures us it all works for them.

For some reason github refuses to show me diffs on the CL you linked to.

Received on Thursday, 23 April 2015 08:46:51 UTC