Re: [IOT] [Standards] Status of XEP-xxxx: Sensor-over-XMPP

Hello Anthony. Great to hear from you. I'm sorry I didn't know of your XEP until now. Our implementations has been one of transport mainly, only the advantages of XMPP transport, including client authentication, and network topology independence. We've left most of meter/sensor communication to the protocol transported on-top (being it M-bus, Modbus, etc.), with payload being identified using a content type attribute. Our desire is to create a better abstract interface which is protocol independent, which would allow for the same things. (And we have such an abstraction already defined for other purposes, but it is not used in our xmpp implementation, so we have some ideas on what we would need in a new XEP.) However, we don't see it as necessarily advantageous to start specifying interfaces for different kinds of devices in the XEP. We believe this is a bad abstraction. (Something outside of the interfaces defined, would not be valid until one updates the XEP.) However, an interoperability section online would probably be required, for people wanting to define specific interfaces for certain use cases. We don't either see a difference between sensors and a actuators in how devices are handled, they are examples of two extremes of the same abstraction: To us most devices consists of a collection of both readable values (i.e. "sensor" values, but also other kind of non-sensor values) and writable values (configurable values, output values, etc.). A PLC is an example where you cannot simply say it's a sensor (it may have sensing capacities and most probably will) or an actuator (it most probably has output values, but might not use them). This we want to include in a XEP (and if previous XEPs could/should/shouldn't be used and why): * Abstract interface description, describing available sensor resources (readable/writable values, data types, meta information, etc.) * Request/response mechanism for readout. * Request/response mechanism for output. * Spontaneous reporting of momentary values based on subscription rules, hysteresis levels, and/or other logic. * Node topology information (if sensor part of larger whole, like a concentrator, for instance). * Abstract metering data description, including: o Timestamps o Description o Units (if numerical) o Precision (if numerical) o Statuses (sequence of flags: error, QoS, tampering, etc.) o Data types (numerical, Boolean, string, date & time, enum, time span) o Value Types (momentary values, historical values, status values, informative values, etc.) o Localization information We also want to include several other things in a separate xep (like control logic), based on what's possible if XMPP is connected to a semantic web environment. Would anybody be interested in working on such a XEP with us? Sincerely, Peter Waher From: Anthony Rowe [mailto:agr@ece.cmu.edu] Sent: den 17 december 2012 13:45 To: mat henshall Cc: gauravbhatia(a)cmu.edu; iot(a)xmpp.org; Joachim Lindborg (joachim.lindborg(a)sust.se); XMPP Standards; marioberges(a)cmu.edu Subject: Re: [Standards] Status of XEP-xxxx: Sensor-over-XMPP Hi all, Great to see so much interested in XMPP for IoT applications. The XEP was basically a summarization of how we had been using XMPP and pubsub in our sensor andrew project. We were also lucky enough to get a fair amount of input form Charles at Google with some interesting use-cases and a few idea on reducing message traffic. The main hangup was that the XMPP council didn't like the idea of linking sensors and actuators and there was some concern if pubsub was generally a good fit for actuation. We basically decided that as researchers in a University environment we would just keep doing what we were doing and revisit the XEP in the future if there was more interest. One easy fix would have been for us to split the XEP into a sensor and actuator document. I would be very interested to hear about how you all have been structuring your XMPP communication. Is it similar to what we proposed? Is anyone using pubsub? We are still actively using our proposed model with a few minor changes to the XEP. Our most recent version of the XEP document can be found here: http://sensor.andrew.cmu.edu/xep/sox-xep.html We also have a simple tutorial for getting up and running with our tools: http://sensor.andrew.cmu.edu:9000/raw-attachment/wiki/quick-start/sensor-an… As for implementations, we have a C, java and python version of our library (called SOX) that we have continued to develop. We have been focusing our recent efforts on the slightly higher-level problems associated with data storage, meta information management and application hosting. These all for the most part exist above the XEP which is just our simple message passing format. I wouldn't be opposed to revamping our XEP with other's inputs and trying to resubmit. We have been getting interest from our corporate sponsors to take another shot at the XEP. -Anthony On Dec 17, 2012, at 11:16 AM, mat henshall <mat@squareconnect.com<mailto:mat@squareconnect.com>> wrote: We are using XMPP for both sensor reporting and control for building and home automation applications. We have implemented a very rich set of stanza's that cover almost all common types of devices and it is designed to work on very low resource embedded devices. This implementation is currently in closed beta although there are some very large brands who have started to develop applications and hardware using our protocol and technology. Our intention is to make the protocol public once we had a full working public available implementation. When we became aware of the proposed XEP extension mentioned here we were already a long way down the road with our own, and as there is so much more to making a complete system than is exposed in this XEP, we felt we needed a working implementation to compare and contrast and make meaningful contributions based on experience... We would be excited to work with others on creating a standard... the problem as always is time to commit to this exercise. That being said, we do have executing code and multiple devices talking to each otehr across continents... so I think we are at the stage where we could add to any serious attempt for standardization. Mat On Mon, Dec 17, 2012 at 9:54 AM, Hannes Tschofenig <hannes.tschofenig@gmx.net<mailto:hannes.tschofenig@gmx.net>> wrote: Hi all, I was actually wondering myself about the status of XMPP & SIP usage for sensors. I dropped Peter a mail a month ago to hear more about the deployment situation. It seems that if there are implementations then they are using HTTP. Ciao Hannes On Dec 17, 2012, at 5:47 PM, Matthew Wild wrote: -- Mat Henshall Founder and CEO, Square Connect, Inc. San Jose, CA www.squareconnect.com<http://www.squareconnect.com/> cell: 650.814.7585