Re: [Tm-rid] 1st steps - time lines and drafts

[Carsten Bormann <[e-mail]>]

On Aug 1, 2019, at 22:55, Michael Richardson <[e-mail]> wrote:
> 
> My experience with CBOR is that unless you want a flat non-extendable
> C-struct, and you will ALWAYS need *ALL* of the fields,  odds are that CBOR
> is smaller. 

Carsten’s compactness principle of data encoding: It is *always* possible to invent something bespoke that is somehow more compact than using an existing data serialization scheme.
(Corollary: It is much harder to invent something that is actually better in more than the compactness dimension in the long run.)

So when the OSCORE people needed to shave off a byte or two from their CBOR-based encoding, they went for a top-level bespoke compression of a structure that continues to use the CBOR data model (see Section 6.3 for that): https://tools.ietf.org/html/rfc8613#section-6.1
This escape hatch is always available.

Apart from that, it *is* possible to come up with CBOR structures that are actually wasteful; I hope there are enough people who have already used CBOR here that this outcome can be avoided.

Grüße, Carsten