Re: [Cbor] Using cddl-control to dissect byte strings

[Carsten Bormann <[e-mail]>]

I think the subject of this message should be 

## Using cddl-control to make CBOR even less schema-less

But indeed, what you propose is one way to compose your byte strings, if you *have* to do it.

I’d probably use .abnfb for many of these; at least if (like in this example) you don't need to go back to CDDL names (which is a shortcoming I have a plan for, but that will happen later this year).

You mentioned the cddl tool generator — does the validator choke on this?

(I don’t want to think about .det here.)

(BTW: I would like to fix one bug in the generics implementation before we aggressively market the use of generics to the CDDL user crowd.)

(BTW^2: Why the [()]?  It does look very important, but maybe not necessary?)

Grüße, Carsten


> On 2021-07-30, at 13:36, Christian Amsüss <[e-mail]> wrote:
> 
> Signed PGP part
> Hi,
> 
> continuing my theme of creatively abusing CDDL constructs, I may have
> found a practical use case for .cat that is not just about constants
> handling.
> 
> (The usual caveats for these tricks apply: While CDDL has the
> expressitivity to do it, common tools WILL PROBABLY NOT implement the
> full genericness).
> 
> In EDHOC, a byte shaving step may require moving two semantically
> separate byte strings, one of which has known length, into a single byte
> string; in a minimal similar example:
> 
>    message = [(number: uint, key: bstr, data: bstr)]
> 
> becomes
> 
>    message = [(number: uint, key_and_data: bstr)]
> 
> Unfortunately, this means parsers built from the CDDL will need an
> additional manual step in which the implementer writes down the prose
> description of key_and_hash into code that splits key from hash.
> 
> With cat, this can be expressed in a way that the CDDL parser generator
> has a chance to do the right thing again:
> 
>    message = [(number: uint, key .cat data)]
>    key = bstr .size 16
>    data = bstr
> 
> As the length depends on earlier exchanged messages, one might use
> generics to pull out the constant of 16 -- probably breaking the tool's
> neck in the process: (Carsten's Ruby generator tool is still with me on
> this)
> 
>    message_alg42 = message<4>
>    message<L> = [(number: uint, key<L> .cat data)]
>    key<L> = bstr .size L
>    data = bstr
> 
> 
> I don't know at this point whether any form of this should be suggested
> for use with EDHOC if and when the byte shaving change is performed.
> 
> Opinions from CBOR?
> 
> Thanks
> Christian
> 
> -- 
> To use raw power is to make yourself infinitely vulnerable to greater powers.
>  -- Bene Gesserit axiom
> 
>