Re: [Doh] [Ext] Proposal to close off these threads

[Tony Finch <[e-mail]>]

Patrick McManus <[e-mail]> wrote:

> it doesn't need to be more complicated. The mti wireformat is defined as
> one message (or if its not - it should be clarified) that might be limited
> to 64KB. I thought one message was already clear and we were debating the
> 64KB point.

I was talking about how to handle query meta-types rather than response
sizes. My point was that there are several reasons an AXFR or IXFR query
can break the single message requirement.

> so if you can't meet those constraints, you can't use that media format.
> but its not inherently about the query type.

I think I would prefer it if the client could predict the response media
type based on the query, rather than the server choosing a media type
based on criteria outside the client's control. This is a programmer
interface so I think predictability is more important than DWIMmery.

> > > "HTTP defines status code 406 for cases where the server cannot generate a
> > > representation suitable for the client". We can probably just say that.
> >
> > My DoH server sends a 200 OK HTTP response containing a NOTIMP DNS
> > response.
>
> Personally, I think you're better off with the HTTP error in this case as
> the problem is with finding a matching HTTP transport.

I don't think that is supported by the current draft text: it explicitly
says that DNS errors should be returned as 200 OK application/dns-message
responses, and there's no other defined media type that we're failing to
negotiate.

Tony.
-- 
f.anthony.n.finch  <[e-mail]>  http://dotat.at/
Trafalgar: Northerly or northwesterly 4 or 5, occasionally 6 in east. Slight
or moderate. Fair. Good.