Re: Discovery API

Hi Dong-Young,

On Jun 9, 2011, at 04:38 , Dong-Young Lee wrote:
> Since a concern was raised regarding device/service discovery in the charter review, I wonder if anybody in the WG has a plan to contribute about discovery API, and if so, whether there’s preference (in this WG or others).
> It would be an important factor in the charter discussion.

I won't discuss the specifics of the concerns that have been raised in charter review since they are (unfortunately) member-confidential and this is a public list (if the companies that expressed these concerns would like to bring them here, it would certainly help the discussion). However, they seem to fall into a few categories which I will try to address:

• Coordination on this topic with other groups is unclear. Whichever group takes on this deliverable, coordination with other groups will be required. We've always done our best to coordinate with other groups, and plan to continue doing so. If the charter is unclear on this point, we should add the relevant groups to it and clarify the situation.

• The domain problem is different from the other APIs that DAP tackles since it is network related rather than device related. Actually, several of the existing DAP deliverables such as Contacts or Calendar are designed to be implementable atop network services, and in fact they are completely agnostic as to what goes on behind the scenes. I would expect a similar approach to be used for a Discovery API. In truth, the term "device APIs" was always a misnomer. A lot of the functionality exposed isn't provided by the device. Some is provided by local services, some by network services. It also often depends on the implementation. What is meant is something like "access to stuff outside of the HTML+HTTP box" but we don't have a nice word for that. Of course, people who participate in the development of the Discovery API will need some domain knowledge — in the same way that we need domain knowledge for Contacts, Calendar, etc. But I'm confident that we will have such people, and what's more we can coordinate for review from expert groups (as was done with CalConnect, vCard, etc.).

• Some comments seem to think that we'll be creating a network-level discovery protocol (due to the many meanings of the term "API"). The idea is rather to produce a Javascript API to expose existing discovery services.

There has been some preliminary work on the topic, which I think shows that contributions are ready to be made, see for instance http://lists.w3.org/Archives/Public/public-device-apis/2011Apr/0068.html.

I hope that this addresses some of your concerns; comments and questions welcome!

-- 
Robin Berjon - http://berjon.com/ - @robinberjon

Received on Thursday, 9 June 2011 09:15:51 UTC