Add a new error code for when the TRR provider is unavailable
Categories
(Core :: Networking: DNS, enhancement, P1)
Tracking
()
People
(Reporter: nhnt11, Assigned: nhnt11)
References
Details
(Whiteboard: [necko-triaged][trr])
The frontend spec calls for two different error pages when the resolver isn't available and when the resolver can't resolve the hostname. We need a way to distinguish these cases. Currently NS_ERROR_UNKNOWN_HOST is used for both.
|
||
Updated•5 years ago
|
|
Assignee | |
Comment 1•5 years ago
|
||
I think, after reading the spec [1], that it's sufficient for this use case to use UNKNOWN_HOST when the HTTP status was 2xx and the new one otherwise.
|
|
Assignee | |
Comment 2•5 years ago
|
||
I investigated this with Valentin's help and it's going to be a pretty significant chunk of work to propagate the distinction between "provider unavailable or misbehaving" and "provider tried and couldn't resolve host". DNS stuff happens separate from the request/channel and it's not trivial to associate the information with the actual http request.
For now, we can probably use the trr mode in the frontend to show a different error page, but we're going to need to add some infrastructure to propagate DoH information out of the DNS code and maybe associate it with the channel or something. I'm tentatively closing this bug for now, we can file new ones when the time comes for the infra I just mentioned.
Description
•