Re: [Cbor] CDDL parsing questions

[Brian E Carpenter <[e-mail]>]

A supplementary question below:

On 18-Aug-22 11:52, Derek Atkins wrote:
> HI,
> 
> On Wed, August 17, 2022 7:17 pm, Toerless Eckert wrote:
>> Sorry, drive-by question (as in: i am not following the mailing list
>> thoroughly, so if this was already discussed, i happily take a pointer
>> and apologize for duplication).
>>
>> When defining data structures via CDDL for protocol headers, parsing
>> of such protocol headers does not only include parsing the CBOR or
>> JSON data structure, but also matching/parsing that data structure
>> into the right CDDL names.
>>
>> CDDL makes it easy to define names that can require quite complex
>> "lookahead" parsing. This is not uncommon in natural or computer
>> languages, but to my memory this is usually painstakingly avoided
>> in "classical" (pre-CBOR) packet header structure designs (at least
>> at network/transport layer). Now, with the abstraction of CDDL,
>> new designers may not even think about the resulting parsing complexity.
>>
>> I could not find anything about this in rfc8610, so i was
>> wondering if the WG ever discussed this and/or if members
>> have thoughts about this. I for example just ran into this
>> with one of our CDDL defined protocols. Wonder if it exists
>> in other CDDL defined protocols too, and if so, with what
>> degree  of complexity.
> 
> My personal feeling is that the receiver needs to know a priori what the
> schema is in order to properly parse the object.  CBOR had a clear parsing
> mechanism (you have arrays and maps), but the semantics of the array
> offset and/or map keys is part of the CDDL schema and not part of the
> parser, per se.
> 
> In other words, you can parse the CBOR into its underlying structure
> without knowing the schema, but in order to process it, you need an (out
> of band) way to know the schema.
> 
> Tagged objects help identify what the object is, but without that, you
> pretty much have to know out-of-band.

And does that also apply if you have some sort of compiler-like parser
driven by the CDDL? Or do you assume that the parser is hand-coded
in order to implement the given CDDL and insert the out-of-band semantic
knowledge?

    Brian

> 
> Just my $0.02 (from having recently designed a CBOR data protocol,
> described by CDDL).
> 
>> Cheers
>>      Toerless
> 
> -derek
>