Re: [Cbor] Multiple questions with CBOR for DNS

[Martine Sophie Lenders <[e-mail]>]

If we make the type mandatory, it is not necessary to define it as 
`undefined` or `null`. Currently, it defaults to 28 (AAAA records), so 
if we make the type necessary it should be the according record type 
number. But that exactly should make clear, that we are not just adding 
a single byte but up to three, as the number space of record types is 1 
to 65535.

That said, maybe it makes sense to make the type necessary. The AAAA 
record type was picked as a default based on what we see on the current 
Internet, where address resolution types (i.e. A and AAAA records) are 
the most prevalent, but that might change. Will look into that.

Best
Martine Lenders

On 09.11.23 19:38, Joe Hildebrand wrote:
> Wouldn't it be easier to make type-spec mandatory for all questions, and allow `undefined` or `null` (pick one) as the type-spec?  You're only adding a single byte.
> 
> —
> Joe Hildebrand
> 
>> On Nov 9, 2023, at 8:25 AM, Carsten Bormann <[e-mail]> wrote:
>>
>> How about:
>>
>> question-section = [
>>   * full-question,
>>   ? last-question
>> ]
>>
>>     full-question = (
>>       name: domain-name,
>>       type-spec,
>>     )
>>
>>     last-question = (
>>       name: domain-name,
>>       ? type-spec,
>>     )
>>
>> It is a bit hard in this form to maintain the requirement for at least one question, because if the last question is also a full question, it will be gobbled up by the `* question` part.
>> (An intersection would work, but requires two parsing threads.)
>>
>>> In the current draft, it is not, but if we go for the name component route, it will become hard to distinguish between the name for a second question or another component of the name for the first. Just necessitating at least a type in-between the names would help with that
>>
>> Grüße, Carsten
>>
>> _______________________________________________
>> CBOR mailing list
>> [e-mail]
>> https://www.ietf.org/mailman/listinfo/cbor
>