2) Even if the above were not true, at that point you're back to an internet enabled smart home device system, and now we're simply picking which vendor to trust over the other. But in both cases, the option for the vendor to collect telemetry data about your usage of the products exists.
There is really no viable way for the typical consumer to be able to both have a good product experience for something like this, and to prevent a cloud vendor from having access to their data. Unless I'm missing something obvious.
Home Assistant Cloud is essentially a TCP-level proxy (IOW Nabu Casa sees jack squat):
> The remote UI encrypts all communication between your browser and your local instance. Encryption is provided by a Let’s Encrypt certificate. Under the hood, your local Home Assistant instance is connected to one of our custom built UI proxy servers. Our UI proxy servers operate at the TCP level and will forward all encrypted data to the local instance.
> Routing is made possible by the Server Name Indication (SNI) extension on the TLS handshake. It contains the information for which hostname an incoming request is destined, and we forward this information to the matching local instance. To be able to route multiple simultaneous requests, all data will be routed via a TCP multiplexer. The local Home Assistant instance will receive the TCP packets, demultiplex them, decrypt them with the SSL certificate and forward them to the HTTP component.
> The source code is available on GitHub:
> SniTun - End-to-End encryption with SNI proxy on top of a TCP multiplexer
> hass-nabucasa - Cloud integration in Home Assistant
Typical consumers have no way of ensuring their UI is, in fact, encrypting the data and not farming it out. They cannot verify the source code themselves, because they don't have the technical skill set they'd need to do so (nor, frankly, the time). They're reliant on the goodwill of whoever packaged and installed the offering for them not doing anything to that offering.
Technical power users can circumvent this because they can build/install from source, verify keychains, read the source, etc. Non-technical users can't do this, and need someone to help them. That someone will most likely be in the form of a third party organization that does this in exchange for money. They're placing their trust in that third party.
The point I'm getting at is that, eventually, a consumer has to trust a third party who may have incentives that don't align with their own. They're just playing a game of which vendor to place that trust in. This is why centralization is still the predominant architecture choice for the overwhelming majority of products, even in a world where myriad decentralized solutions exist for almost everything. It turns out that having bespoke third parties run decentralized solutions for customers is often not a better product experience, and still has the same root problem even if it manifests in different ways.
