Re: [fdt] [Webtransport] Standards for protocol headers?

[Carsten Bormann <[e-mail]>]

On 2020-06-24, at 04:27, Scott Morgan <[e-mail]> wrote:
> 
> Thoughts?

Well, you seem to want to invent another lexical syntax.

2020, we don’t do that any more.

For things that are text and need to be exchanged as text, JSON is the lexical syntax now.
Where you need more flexibility (e.g., include JPEG) and don’t have a need for the encoding to be text, use CBOR.

This turns the question “what is the lexical syntax” (something which you would express in ABNF) into the question “what is the structure of the data” (something that you express in CDDL).  For a theoretician, of course, both are syntaxes, but there is a lot of benefit in handling the lexical syntax with a known good software component (many bugs that turn into attack vectors are in parsers for lexical syntax).

[That doesn’t mean you cannot have fatal bugs in the component that handles the structure of the data — we have just reduced the attack surface a lot, and we can use an optimized component for the lexical work.]

This doesn’t mean there isn’t a place for highly refined encodings such as HDF5, or legacy structural encodings such as RIFF (underlying WAV etc.) or JPEG’s byte stuffing based delimiters.  There are even new legacy encodings of this kind being defined :-) (e.g., Extensible Binary Meta Language, draft-ietf-cellar-ebml).  But for 98 % of all new work, JSON and CBOR are it.

Grüße, Carsten