But this 1) adds a technical hurdle, 2) reduces accuracy of the data captured, and 3) adds a potential legal barrier, as the data is not theirs anymore
(before ISPs could claim that the use of their DNS gave them right to use the data collected)
ESNI can solve this, by encrypting the SNI header while still allowing serving multiple https domains on one IP. But it’s currently mainly implemented by cloudflare.
The benefit they do provide is authentication (ensure that google.com is really google.com) and protection against man-in-middle.
And those are great benefits!!
But it doesn't help anyone to misstate and/oversell the benefits of a new service/protocol.
Yes, but your ISP (including "that coffee shop whose wifi you're using" in ISP here) can't necessarily tell which hostnames you are resolving, if you use DoH (subject to some limitations about what happens after you've resolved the hostname). This is in fact a privacy feature
This is why ESNI is being rolled out, yes?
Is there anything in DoH mitigating this? Or maybe is this attack vector negligible in practice because most servers typically host multiple websites? At least this adds plausible deniability in a wide range of situations I guess. But is it really true? And for example, would it really help in a country where a website like Facebook is censored? (Since their IP addresses are dedicated).