Re: [PATCH 2/2] Define ENONAMESERVICE and ENAMEUNKNOWN to indicate name service errors

[David Howells <dhowells-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>]

Jim Rees <rees-63aXycvo3TyHXe+LvDLADg@public.gmane.org> wrote:

>   > Would this be the same as NXDOMAIN?  That is, does it mean the name server
>   > couldn't find a record, or does it mean that the record doesn't exist?
>   
>   Is there a way to tell the difference?  Can you store a negative record in
>   the DNS?  Or is it that the DNS has records for the name, just not records
>   of the type you're looking for (eg. NO_ADDRESS/NO_DATA from
>   gethostbyname())?
> 
> It's an important distinction to the resolver if you want to avoid dns
> hijacking.  See rfc2308.  There doesn't seem to be a way to tell the
> difference from the gethostbyname call, which was designed before this was a
> problem.  The on-the-wire dns query protocol does make the distinction.
> 
> I suspect kernel dns clients won't need to know the difference, but I think
> it's useful if we decide on and document the meaning of the error codes.
> Maybe the answer is that ENAMEUNKNOWN means the same as a HOST_NOT_FOUND
> from gethostbyname().

Should I propose an extra error code?  Perhaps giving:

	ENONAMESERVICE	"Network name service unavailable"
	ENAMEUNKNOWN	"Network name not known"
	ENONAMERECORD	"Network name query returned no records"

Note that ENONAMESERVICE covers all of: not having a name service configured,
not being able to contact the configured name server and the configured name
server not being able to chain to the authoritative name server.  However, I
think this is probably okay.

David