Re: Supporting OCF resources from the web of things

Hi Zoltan.

Thanks for the kind words and feedback.

> On 5 Dec 2016, at 19:24, Kis, Zoltan <zoltan.kis@intel.com> wrote:
> 
> Hi Dave,
> 
> This is excellent work, and json is indeed more readable. However it seems to miss some information, like interfaces supported, supported methods for a resource/thing, error cases etc.

The supported methods are defined in RAML and generated from the thing description. I found that for the OIC 1.1 resources, it was possible to infer the methods from metadata on whether properties are writeable or not.  I would appreciate some help with gaining a better understanding of the OCF error cases as these don’t seem to be covered in RAML, but are perhaps specified elsewhere.  We need a better understanding of how to handle errors for the APIs we want to standardise for the Web of things, and the extent to which these can abstract over the specific errors for different IoT platforms.

> I wonder how to model in WoT the notion and usage of different OCF interfaces  (baseline, linked list, batch, etc), collections, scenes, etc. when using REST methods. 

It is not obvious to me that these need to be exposed to Web applications in all their detail. The OCF link schema, for instance, has large number of properties, many of which are specific to the OCF framework. A driver for interoperating with OCF devices would know how do deal with these.  The aim is to avoid the need for application developers to have to work with IoT platform and protocol specific details for the devices they want to interact with, given that there are many IoT platforms and protocols.

> I think it may be simpler to first distill the intended OCF use cases, then reapply them in WoT, rather than just focus on translating from one semantics to another (which you proved is possible for the presented resource cases).

The purpose of re-modelling the OCF resources was to explore the minimal metadata requirements necessary for interacting with OCF resources. I can imagine a business case for applications that exposes OCF resources using a different model, perhaps even a composite model including a mix of OCF and non-OCF devices. However, behind this, there would remain a need to interoperate with the OCF devices, and the consequent need for modelling the resources that OCF exposes.

Best regards,
—
   Dave Raggett <dsr@w3.org <mailto:dsr@w3.org>>