Re: [Cbor] I-D Action: draft-ietf-cbor-network-addresses-04.txt

On 26-Apr-21 03:30, Carsten Bormann wrote:
>> - for 261 I would at least include those arguments and make it clear when to use or the other
> 
> Sounds good to me.
> 
>> - so you don’t want to support the commonly used shortcut of specifying both and address and a prefix in one?
> 
> That is a valid point.
> But I think the data should indicate whether they are just a prefix or an address and a prefix.
> 
>> - in ip agnostic code I often find I just need an IP address instead of having to care if it’s 4 or 6. It would therefore be nice with a general IP address and ip prefix type too. 

This is very dependent on the context where you are using the address. Sometimes you need to be address-family independent and sometimes not. It can also be programming-language dependent. And it can be API dependent (who are you passing the address to?). So I think this will have to be handled by the parser of the decoded CBOR, not by the CBOR tagging.

> If I understand this right, we already have that.  One type with two different tags…

Yes, and how you cast it into a particular language's type(s) is not a CBOR issue.

> (CBOR doesn’t have “types” [except for the “major types” on the representation level].  You are free to construct your own types in your implementation, e.g. construct a “Boolean” type from “true” and “false”.  In your case, the “IP address and prefix” type would include values with either tag 52 or 54.  We could define recommended CDDL types, like we did in Section 5 of RFC 8746...)

Sure. But in GRASP we ended up with 
  ipv4-address = bytes .size 4
  ipv6-address = bytes .size 16
which wouldn't suit every case. 

Regards
    Brian

> 
> Grüße, Carsten
> 
> --------------------------------------------------------------------
> IETF IPv6 working group mailing list
> ipv6@ietf.org
> Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
> --------------------------------------------------------------------
> 