On Mon, Feb 25, 2019 at 12:01 PM Carsten Bormann <cabo@xxxxxxx> wrote:
On Feb 25, 2019, at 17:04, Michael Richardson <mcr+ietf@xxxxxxxxxxxx> wrote:
>
>> So my response to Michael is that you don't want your version numbers in the .
>> well-known path, you want them in the DNS records because then they can be
>> used to select the host to connect to according to capabilities.
>
> Interesting. I think that I agree.
For CoAP, we don’t want to become dependent on DNS. But we can do similar things with discovery and “rt” target attributes.
But what do you mean by DNS?
There is the DNS name registry, the DNS data model and the DNS resolution protocol. All three can be called DNS.
We discussed this at the naming workshop. I see two different types of name, universal so that we all interpret them the same way and local. If I am talking to an Internet of Thing in my house 'open the garage door' is unambiguous. If I am on the Internet, it is not.
We can certainly have a personal name resolution scheme and this should be a standard. But we should not introduce a new data model unless we really need to.
The DNS client-resolver resolution protocol is awful. Lets stop trying to put lipstick on that pig and sidegrade to an alternative resolution protocol that preserves the same data model.
One of the reasons I want to establish the convention of resolving the service named 'foo' using the convention it maps to /.well-known/srv/foo/ is precisely so that in protocols such as coap, we can make the service name a transport connection attribute and dispense with .well-known entirely.
So putting it together, lets say Carsten has two devices in his house, that support the not-on-fire protocol that is offered over HTTP, CoAP and in one room, Splunge:
_not-on-fire._tcp.carsten.house TXT "trans=coap trans=http"
_not-on-fire._tcp..carsten.house SRV 0 10 80 garage.carsten.house
_not-on-fire._tcp.carsten.house SRV 0 10 80 kitchen.carsten.house
_not-on-fire._tcp.
kitchen.carsten.house TXT "trans=splunge"
The development tools I use allow me to offer XML, ASN.1, JSON, etc encodings of the same service very easily. It is just a switch on the compiler right now. Next generation it will be just a different library.
On the naming side, I have a scheme for that as well. But that takes a rather more holistic view of the problem. My basic idea is that if I buy a device, I should be in control of it and it should only interact with me and other devices I specify for it to interact with. It should only use the name services that I nominate and no others.
