12 Oct
2016
12 Oct
'16
3:11 a.m.
Hello
I propose that we form a special interest group (see
XEP-0002) regarding the
use of XMPP in the Internet of Things:
http://xmpp.org/extensions/inbox/iot-sig.html
The IoT SIG sounds like an excellent idea.
Firstly, it'd be useful to gather a sense of the
current state of
play. It seems to me we have a number of IoT-related XEPs and
proposals - due to a huge amount of effort by Peter Waher - but its
not clear to me which of these have any traction. It would be great if
people working on IoT (and using XMPP) could say which of these are
generally working well for them.
It might be of interest to you, to know that the IEEE is working to form a new Group on
”IoT Harmonization”, in which XMPP plays an important role. XEPs under consideration at
this forming stage are 0323, 0324, 0325, 0326 & 0347. You can review the slides from a
presentation I held yesterday on this topic:
http://www.slideshare.net/peterwaher/iot-harmonization-using-xmpp
Let me know if anybody working with XMPP & IoT is interested to participate in such a
working group. The goal is to form the group in January and have a first draft ready for
balloting by the end of 2017.
Secondly, I'm of the opinion - and opinions can
always be changed -
that the existing IoT proposals are something of an isolated suite.
Yes and no. They have been abstracted. Some are more IoT related (323,324,325,326,347),
others are more generic. EXI was written as reaction to the observation that XMPP is too
verbose for some IoT applications and some networks. The Event logging (0337) is a generic
infrastructural need, but it arose from logging in distributed IoT systems where many
clients lack displays. HTTP over XMPP (0332) arose from the need to define web queries
among distributed sets of sensors. Dynamic Forms (0336) as a way to create richer data
forms, but is used in IoT since data values might take some time to fetch, and fields need
to be able to be updated dynamically. Form signatures (0348) is needed to automate the
creation of XMPP accounts, in a secure manner. Others that wait approval are also written
to be generic, such as the QoS proposal – which originated as a need from IoT but has
generic value. Event subscription is more directly IoT-related.
Looking at the IETF MILE Working Group, we have the
XMPP-Grid proposal
which seems a similar shape to the IoT proposal, and similarly uses
little of the existing mechanics we have. For example, it provides a
publish-subscribe facility, a registration facility, and so on. The
payloads are different, but the essential goals the same. I cannot see
what would drive a difference in the containing protocol between (say)
counts of stanzas in an XMPP server, temperature readings in a sensor,
and sightings of a Cyber Observable pattern.
There’s already a publish-subscribe based IoT solution defined by the UPnP, and published
by the OIC/OCF:
https://openconnectivity.org/resources/specifications/upnp/iot
To create another one for this purpose seems unnecessary.
While publish/subscribe might work well for several use cases, it’s not sufficient to
cover all use cases, not even the most important use cases. It’s sole purpose is to make
mass-dissemination more efficient. The IoT XEPs however, allow for a more general
architecture, which is not limited to a single communication pattern, but allows for most
patterns used today, depending on what is to be accomplished. See my presentation above
for more information.
Best regards,
Peter Waher