Good suggestion, but where and how does HA receive callbacks? I would guess that almost all HA instances are behind residential LANs and most aren't accessible on the public internet. You could use dynamic DNS and forward ports, but that's flaky, you might run into CGNAT, etc. And anyway, it's best if your HA instance isn't publicly addressable; mine is only accessible over my personal WireGuard VPN and I intend to keep it that way.
I'm sure this is a solvable and solved problem, but I do believe it is non-trivial, and potentially a major headache for a company to implement just to support a tiny niche of users. I'd be delighted to find out I'm wrong though!
And, unfortunately, the business case isn't there, since this weakens lock-in effects. I don't endorse this reason—that's why I run my own HA instance and don't buy or use any products that require the cloud or otherwise can't be operated entirely locally (including flashing Valetudo to my robot vacuum!).
If you pay for the home assistant cloud subscription (built into HA, ~5 USD/mo) they can provision custom callback URLs for you so you don’t have to expose your HA instance. I have this setup for certain integrations such as Samsung Smart Things.
It’s not a perfect solution since it costs money but it’s a nice alternative to exposing your HA instance or some other front end proxy to the internet.
Unfortunately it's not actually that different in effect -- Nabu Casa proxy the encrypted TCP connection, rather than terminating TLS and proxying HTTP, which is great for privacy but not so much for providing an extra layer of security on top of HA itself.
It is also much easier for those without easy access to extra static IP addresses. Given the target audience I think it's probably the right approach.
I don't think it's entirely devoid of security improvements---you need to know the webhook address in order to get access to talk to a HA instance which would be a lot more difficult than just port scanning for an open (perhaps unpatched) HA instance on the open internet. I would still prefer it though if things would expose a local API or speak MQTT however.
Open a TCP connection from the instance to the cloud service. I don't know about all consumer routers, but I just checked mine and the default TCP established timeout is 7440 seconds. Idle timeouts are supposed to be at least 2 hours.
If you served the entire US (130 million households) and had a 1 hour keepalive, that's only 36k packets per second, which is nothing.
You could also auto-train the idle timeout by using a pair of TCP connections. One uses a known good value while the other probes upwards until it finds its connections start getting closed (with some optional binary search fanciness), feeding new known good values back to the first.
MQTT is the solution for this. Note that the garage door openers talk MQTT to the myq service (over TLS with preshared keys). It should be possible to subscribe to events from your garage door opener(s) and also to send commands to it.
but MQTT alone doesn't solve the challenge for some Internet server to push messages to a Home Assistance instance running inside a home network / behind a router / behind a firewall / NAT unless a port is opened on the router, or long-polling is used.
I'm sure this is a solvable and solved problem, but I do believe it is non-trivial, and potentially a major headache for a company to implement just to support a tiny niche of users. I'd be delighted to find out I'm wrong though!
And, unfortunately, the business case isn't there, since this weakens lock-in effects. I don't endorse this reason—that's why I run my own HA instance and don't buy or use any products that require the cloud or otherwise can't be operated entirely locally (including flashing Valetudo to my robot vacuum!).