On Sun, 18 Dec 2016 21:47:17 +0100 Florian Schmaus <flo@geekplace.eu> wrote:
I don't think that XMPP has an inherent overhead besides it being XML based. And that overhead can likely be heavily reduced by EXI and the like.
The main problem, and that is not specific to XMPP, is bad protocol design which imposes unnecessary packets (in XMPPs' case: stanzas) on the wire.
Probably also the point that you have indicated needs to be addressed, but for my personal experience, what people is indicating generally as one of the main issues of XMPP compared with other IoT protocols is related to the use of XML. But I agree that this issue can be addressed with the integration of EXI like proposed in XEP-0322.
Of course. No one is talking about forbidding it. But symmetric presence subscription should not be be *required*. And I think we should design the XMPP IoT protocol(s) based on that assumption.
Totally agree.
Could you elaborate on those?
Sure, I think that the presence mechanism is one of the most important feature that XMPP can provide in the IoT scenario. If every "Thing" is associated with a XMPP JID, its presence can be used to know in real-time if it is online or not and possibly more specific info, using the status. As I said, this feature can be useful in several scenario: - smart-factory, where the presences can be used to keep monitored the devices in one ore more industrial plants, - smart-home, where the same feature can be used for the equiments and sensors deployed in an home. - smart city, where this feature can be used to monitor the status of smart objects spread in the city scenario. More in general, every time can be useful to know in real-time if one device (or virtual entity) is available or not and take a decision based on this information. I hope I made myself clear, Best Regards, Davide