ESP8266 Community Forum

- Thu Nov 15, 2018 12:21 pm #79164 I've come up with an unusual way of controlling a Sonoff smart switch without needing to use any Wifi access point. It uses Wifi probe request messages so neither the client or the Sonoff needs to be connected to an access point - the client sends probe requests with its mac address encoding which Sonoff and whether to switch on or off.

Wondered if anyone can see any flaws in this approach?

Details here: https://github.com/HarringayMakerSpace/sonoff-adhoc

- Fri Nov 16, 2018 4:03 am #79171 At first I thought: "What a great idea, turning a Sonoff ON when device Z (like my phone, which I always carry around) is in the vicinity and turn it OFF again automagically if it isn't seen anymore for an X amount of time".
But then I read you've got an entirely different idea and you don't want to use a network and the idea is to "Misuse" unused MAC-address ranges to control stuff instead.

Personally I don't think this is not how WiFi should work, I even don't see how this would solve a (to me) non-existing problem.
Nobody is forcing you to use a network; an ESP and Pi are perfectly capable in creating an ad-hoc connection when needed.

But then again: I've got an always working (and reachable) network at home I can connect things to, so never had the desire to use WiFi in another way than out of the designed scope.

If you're certain MAC address ranges won't clash, in theory I don't see much trouble.
However I do recall that Apple devices tend to output random MAC address when probing, though I don't now how random these are in reality.

Assumption is the mother of all f*ckups. At least: that's what I'm assuming.

- Fri Nov 16, 2018 4:20 am #79172 Thanks for the comments QuickFix. You say:

... an ESP and Pi are perfectly capable in creating an ad-hoc connection when needed

How would you do that, simply, remembering there can be multiple clients and multiple sonoff's?

- Fri Nov 16, 2018 8:39 am #79179 I've never really thought about something like a no central (access) point problem you're trying to solve.
My gut feeling says to create a special layer around the transport layer inside the OSI model or perhaps just sending UDP datagrams to anyone that wants to listen (and act accordingly, finalizing with an "I'm done" ack).

Assumption is the mother of all f*ckups. At least: that's what I'm assuming.