We really should not be using TXT here because any record that is specific to a protocol needs to have a prefix so that a query can be made for just the records relevant to that protocol. Otherwise the DNS UDP response limit is quickly exceeded. I much prefer to go for a text encoded tag value pair approach for application level attributes because that allows the same syntax to be used in the HTTP (or whatever) header and the DNS record. More generally though, I think that if we are going to introduce a new record it should be a record that allows for more than just HTTP version agility which is essentially one bit of information. Making a UDP round trip for a single bit of information seems excessive. I would rather see any DNS work as being an application layer proposal that should support all application transports and in particular allow negotiation between HTTP, HTTPS and COAP transports. On Mon, Feb 11, 2013 at 10:25 AM, Eliot Lear <lear@cisco.com> wrote: > [Note I'm on vacation and avoiding the computer] > > On 2/11/13 6:17 AM, Mark Nottingham wrote: > > Eliot - a major portion of the conversation was about whether this would > be in a new record type vs. in a TXT record. Do you plan to address that in > your next draft? > > > > I think the conversation was about whether to a use a text formed record > rather than a binary encoded one. Using TXT wouldn't be a good idea for > any number of reasons. > > Eliot > > -- Website: http://hallambaker.com/
Received on Monday, 11 February 2013 15:54:39 UTC