This should be fairly easy to implement. It can and should use existing DNS facilities in the OS; no sneaky encrypted DNS DNS circumvention. It’s almost a free (for the devs) feature.
Windows Server’s Remote Desktop Services’ RemoteApps has a method for discovering RDS servers by just entering the user’s email address in a box, then it sets up the server and proceeds to sign on with the same email address.
For the latter, it just carries over the email address entered-no trick. You can still change if you want. For the former though, it takes the FQDN of the address and in it, it searches for value set on the _msradc DNS TXT record.
I was thinking, ZeroTier should implement something like that, while scrolling endlessly looking for Telegram, where I had texted myself a network ID I needed.
This would allow for each FQDN to “host” or store 1 network ID. For a slightly more complex approach, an arbitrary separator could be used to allow more than 1 network ID per domain, such as sales@sub.domain.tld; where sales would be something like the _ztn-sales TXT record on sub.domain.tld.
The “_ztn-” is to avoid conflict with other records.
Or maybe even go a step further, and mix both approaches; e.g; if the domain hosts a single network ID, have it predefined in the protocol spec lemme try again: in the client, if a domain is entered alone without any separator character in it, look up the predefined record, otherwise perform a custom query. Yep, much better. And of course it would still accept network IDs.
Since you’re targeting IoT, it seems, this could even be useful to mass-redirect clients in the future, since their client IDs could be preempted in the new network. There already is support for this.
I mean… I’m just sayin’. 