Re: A mechanism to encode HTTP version information in DNS

[Amos Jeffries <[e-mail]>]

On 18/02/2013 11:05 a.m., Phillip Hallam-Baker wrote:
>
>
> On Sun, Feb 17, 2013 at 4:31 PM, Adrien W. de Croy <[e-mail] 
> <mailto:[e-mail]>> wrote:
>
>     I've always thought one of the issues with SRV is the transport is
>     in the wrong part - the name, so you can't specify that with the
>     result.
>     For something new, we should look at putting the transport into
>     the RR data instead of the name
>     then it makes it possible to deploy over multiple transports later
>     on, e.g. SCTP, UDP whatever if desired.
>
>
> That is exactly what the URI RR does, the target is a URI which can of 
> course specify a new scheme.
>
> I think it best to treat _http._tcp as simply a code point 
> corresponding to the http URI scheme and the _tcp part as just a default..
>
> If you wanted to specify a different transport then you would either 
> specify a different scheme or you would specify it as a discovery 
> parameter for that scheme.
>

Or since HTTP is effectively a transport in these respects we have _http 
defined alongside _tcp, and make the under-layer transport an option in 
URI RR

   _http.example.com <http://tcp.example.com>  URI 10 10 
"http://www1.example.com/ TCP HTTP/2.0"

With default option for under-layer being TCP when no other is mentioned 
specifically.

Amos