https://www.theverge.com/2017/8/3/16083996/internet-of-shit-...
Or are you saying it's prescient because Apple saw that such capabilities were not economically viable in the long run, and so it was prescient of them to never develop the functionality in the first place?
If all the accessory vendors were to go under and Apple were to stop supporting HomeKit tomorrow, HomeKit accessories and hubs would still operate just fine and alternatives could be created. Even using Alexa or Google Nest for voice control.
From what I can tell, Homebridge implements Homekit as a device, receiving commands. But can anyone build an app or device to send commands, outside of iOS?
The fact that I can't find an Homekit control app for Android strongly indicates this is not possible.
A protocol being fully documented does not mean you can get a device you cannot control (like an Homekit lightbulb) to connect to your server instead of the manufacturer's. If requests are signed and/or encrypted, knowing the protocol won't help if you don't have the keys. See also Tivoization.
> For example, search the web for ‘homebridge alexa’
I had found it. I don't think you realize that there's no Homekit there at all.
The point of Homebridge is to translate from a Homekit-speaking control app/assistant to a non-Homekit-enabled device:
https://blog.theodo.com/wp-content/uploads/2017/08/workflow....
But when you use homebridge-alexa, you're replacing that first link with another protocol (specifically, MQTT+JSON, which is what it uses to connect to https://www.homebridge.ca/).
So you've actually removed all Homekit parts from your smart home system!
It's just terrible, barely working, shoehorned in buggy drivers, "shipped as soon as it can appear functional" code.
You could take a microcontroller, write something simple, and have something that worked to a few 9s.
Instead, IoT vendors spend their time optimizing their data collection and monetization platforms.
"You had one job..." would be a good epitaph for the entire industry.
Not saying the OP isn't having issues, but there might be a sample bias when reading the comments.
Not all smart devices are created alike. Research your purchases.
So far the one I’ve been most happy with is IKEA Trådfri.
It costs roughly half of the comparable Phillips hue offering, but still delivers on any aspect I care about.
Works 100% locally, uses standard Zigbee (non-WiFi) networking. No cloud required.
Supports HomeKit, Alexa, google assistant, not to mention Home Assistant.
More importantly, it supports other vendor’s Zigbee devices (like Hue) and supports working with other Zigbee controllers than IKEA’s own.
Also certified non-shit by kernel-hacker Matthew Garret[1].
Nobody can shut me out of this purchase. This is actually my hardware.
Baby starts moving in bedroom, light bulb flashes in kitchen.
Sonos stopped supporting remote control, no problem, map IKEA remote controller to start/stop Sonos playback.
https://www.zigbee2mqtt.io/information/supported_devices.htm...
https://www.ikea.com/gb/en/products/lighting/smart-lighting/
I also have two Nests. Stringify has now shut down (and was buggy/difficult to use), and IFTTT is now losing access to Nest it seems, so I guess that's a real knockback of at least several years for smarthome.
Even when it was all in theory working, the most simple thing was not actually possible - Nest shows you humidity, so when humidity is over 60%, turn on the dehumidifier (I have one upstairs, one downstairs). Such a simple task coudln't be achieved because of about individual problems, and this to me puts the date of solving it back quite a way.
Edit: They do support UPnP, but still they are wide open to attack.
- security
- support
- safety
With those devices. Don’t get me wrong, I love AliExpress - but I’d be worried about anything running at mains voltage, especially when I’m away from home.
It is hard to make a lasting business off that, which is why everyone wants to do subscription services.
Maybe it's because people trying to get rich quick playing the VC lottery are sucking the oxygen out for people who want to just do a honest to God, "payment in exchange for providing value" type of business.
That implies stuff like "there is one way to do it", "95+% branch coverage in unit tests", and "100,000-1M hours of automated stress testing of real hardware before shipping a new board rev."