A lot of the worst IoT vulnerabilities in the past have been due to exactly that. 'Local' unfortunately isn't something decided at design time, it's decided when someone connects it to a network. Most people plugging these devices in don't have any clue how to simultaneously secure them and connect them to the internet, so they often end up directly on the internet with default credentials or with outdated vulnerable software and a port open. That's the biggest reason all of the major players now just close all inbound ports and reach outbound to a cloud service. It checks both boxes of usability and network security with even the most misguided user.
Yes, this arrangement sucks for people who know better. But we aren't the people in the user stories.
Even though I advocate for a local interface, I also completely agree with your statement.
But, the alternative is we either accept this completely upside down API traffic ratio with locally focused integrations (bad, costs lots of money) or allow a local interface.
Another potential workaround I advocated for was a "cloud down" message that could enable the local interface for those that ONLY go looking for how to do it.
I'd be 100% on board with local control being a minor PITA to enable, as long as it's allowed and supported. I'd even be ok all the way up to needing to reflash the board to allow it.
Making it so that only people who care and are more likely to use it in a responsible way have access is pretty reasonable! Not having the option isn't (in my opinion).
With my business hat on, I'm not so sure. "Why are we spending development resources on a feature that isn't valuable for our target users?"
I could definitely see doing this if it were a product with a prosumer-type angle to the value prop. But for the devices we see on the shelf at a big-box store, I don't think those companies' management considers that valuable.
All companies need to do to support home assistant in this front is:
1. Making sure their app can control the smart devices even while the internet connection is out as long as the phone and the devices is connected to the same lan (local control). Local control adds resiliency to your product, which increase user satisfaction. Don't see it as spending an effort to support home assistant, but instead, see it as making your own product more resilient to unstable internet connection.
2. If your device don't support ZigBee (or other local protocol) and only supports wifi, have the local control api secured with a key. This key will be generated during initial setup and should be retrievable from your app.
That's it. If your devices are popular enough, someone will poke around, see the device has local control api secured with a key that can be retrieved from the official app, and publish an open source integration on HACS. You spent zero effort to directly support home assistant but your users now has an option to use their devices with home assistant and will likely to be a repeat customers.
> 'Local' unfortunately isn't something decided at design time, it's decided when someone connects it to a network.
It's obviously connected to the public internet when it talks to cloud servers, and that's somehow (claimed to be) secure.
Comparing a good cloud API with a poorly designed local API is a false dichotomy. Would you set up your cloud servers with default credentials of admin:admin?
Have a hidden physical switch that toggles local control, and require a physical button press to (re-)generate secure credentials. Have the user upload TLS certificates (non-optional), then hand over the credentials over a secure connection. There, the security of local API should now be up to par with the cloud connection.
There's a at least a dozen ways for set up a secure local API. Whether it is possible was never the question. The question is whether Joe Blow can pick one up off the shelf at Best Buy and successfully implement it. The answer to that question is no. Joe Blow wants to download the app, click the button, and make it work. This is 99.999% of the users that buy something off the shelf at the big box store.
Asking why a Haier dishwasher doesn't have a local API is like asking why a Toyota Sienna doesn't have configurable launch control, power-take-off, or a fifth-wheel. The target market segment isn't looking for those features.
> Joe Blow wants to download the app, click the button, and make it work. This is 99.999% of the users that buy something off the shelf at the big box store.
And I don't dispute that, this option should remain available. What I dispute is the idea that the lack of local control is somehow beneficial to the end user by "protecting" them from vulnerabilities.
The only thing such arrangement is protecting is the manufacturer's bottom line, by allowing them to 1. harvest and sell data, 2. take away features or start pushing upsells when they need to boost their quarterly profits.
> Asking why a Haier dishwasher doesn't have a local API is like asking why a Toyota Sienna doesn't have configurable launch control, power-take-off, or a fifth-wheel.
Well that's just ridiculous, all of those features have significant per-unit cost to implement. Exposing some form of local control would take, if we're being generous, a couple of person-weeks of effort and it would cover the entire product line with a marginal per-unit cost of a single switch.
