Re: Representative sample of industry protocols

Hi,

I think one way would indeed be to prioritize the work on HTTP and
Websockets as we were suggesting in the Web Thing model (
http://model.webofthings.io): HTTP because it is simply the ubiquitous
protocol of the Web, Websocket because it represents a way to deal with the
event-driven real-world supported by a very large number of clients (and
servers). We use Websocket to that aim for years now, we also use MQTT over
Websocket which is pretty easy to achieve and can happen all in the browser
(as both protocols use TCP). In terms of understanding the content of WS
frames there is a standard way of doing so using the Websocket subprotocol
field (see https://www.iana.org/assignments/websocket/websocket.xml).

Then of course JSON is the interop data format on the Web with the ability
through content-negotiation to use a binary protocol (e.g., messagepack,
etc.) and the open door to the Semantic Web via  JSON-LD extensions (
https://www.w3.org/Submission/2015/SUBM-wot-model-20150824/#semantic-extensions)
but I would not make it mandatory: there is a lot of interoperability value
in supporting plain old JSON with a basic agreed upon model.

I think this a trend you can observe in many places. Back 10 years ago not
that many Things protocols were considering the Internet, let alone the
Web. Today however things have changed. Weave is building on HTTP and JSON,
homekit likewise, EnOcean support HTTP at the gateway level, Bluetooth has
a GATT REST API and even Bacnet apparently will support RESTful services:
the IoT is finally getting on the Internet and Web protocols seems to be
the place where the convergence happens defacto, creating a uniform
application layer. However, the semantics of interactions, resources and
payloads structure is not uniform yet. This to me is the Web of Things and
where this group should contribute.

As a side note: the role of HTTP/2 in the IoT for me is also important to
call out: and HTTP/2 will be much more suitable for embedded devices
brining some of the goodness of protocols like CoAP and MQTT to a larger
Web: header compression, binary protocol, serverpush, multiplexing (see
e.g., http://webofthings.org/2015/10/25/http2-for-internet-of-things-1/).

Dom


On Wed, Oct 26, 2016 at 9:50 AM Dave Raggett <dsr@w3.org> wrote:

> Hi Ben,
>
>
>
> I am hearing strong agreement about the value of HTTP as a very popular
> Internet protocol, but not so much about the impact of different
> application domain requirements on the communication patterns. HTTP itself
> can be used in many different ways, and this can lead to interoperability
> challenges. It thus makes sense to identify design patterns for common sets
> of requirements based upon an agreed set of use cases. We can then define
> the metadata vocabulary for declaring how a particular platform is using
> the protocol, as a means to enable interoperability. The Interest Group has
> already done quite a bit of work on this, albeit on a restricted set of use
> cases.
>
> Whilst we can prioritize work on HTTP, we shouldn’t preclude work on other
> protocols, as according to the level of interest amongst the group
> participants. The Interest Group, for instance, has worked on CoAP.
>
> In respect to WebSockets, people tend to roll their own (proprietary)
> protocol using JSON messages. Interoperability would require work on
> standards for these messages. This seems like something that needs further
> incubation to ensure the appropriate level of critical review.
>
> p.s. this is of course just my personal opinion.
>
> —
>    Dave Raggett <dsr@w3.org>
>
>
>
> --
--
Dominique Guinard, Ph.D. ////
co-founder & chief technology officer
+44 79 5153 2987 // w evrythng.com
t @domguinard // w guinard.org
b webofthings.org

About EVRYTHNG: http://bit.ly/smarterIoT
Book: Building the Web of Things: http://bit.ly/wotbook
Bloomberg Innovation Leader 2016: http://bit.ly/1OHR7k7
RedHerring Top 100 2016: http://bit.ly/1WbIF4t
10-billion Products Born Digital: http://bit.ly/1SUHiSN