Re: [Cbor] [Technical Errata Reported] RFC8610 (6527)

[Carsten Bormann <[e-mail]>]

On 2021-04-11, at 18:51, Carsten Bormann <[e-mail]> wrote:
> 
> However, there is one omission in the ABNF: the \u syntax, which is somewhat complicated in JSON because it is followed either by 4 hex digits that are not in the range d800 to dfff or by 4 hex digits in the range d800 to dbff, another \u, and four more hex digits dc00 to dfff.
> Someone needs to sit down and write up the ABNF for that (or find some ABNF that already has done the work).

I was too lazy to find some ABNF so I wrote my own.

76c76,84
< SESC = "\" (%x20-7E / %x80-10FFFD)
---
> 
> SESC = "\" ( %x22 / %x2F / %x5C / %x62 / %x66 / %x6E / %x72 / %x74 /
>              (%x75 hexchar) )
> 
> hexchar = ((DIGIT / "A"/"B"/"C" / "E"/"F") 3HEXDIG) /
>            ("D"
>             (( %x30-37 2HEXDIG ) /
>              (("8"/"9"/"A"/"B") 2HEXDIG "\" %x75 "D"
>               ("C"/"D"/"E"/"F") 2HEXDIG )))
79c87
< BCHAR = %x20-26 / %x28-5B / %x5D-10FFFD / SESC / CRLF
---
> BCHAR = %x20-26 / %x28-5B / %x5D-10FFFD / SESC / "\'" / CRLF


The second change is necessary as SESC is now narrowly restricted to the escape combinations that JSON allows in text strings, and \' is not among those.

You can play with this changed grammar in cddlc version 0.0.3 (`gem update` if needed).

> I’m not sure this update should all be put into an errata item; maybe we should pursue writing this up in a document that updates RFC 8610 that could then also add a less unwieldy syntax for non-BMP code points such as the ubiquitous \u{…}.

I’m still not entirely sure we want to handle this as a bog-standard errata item.
Opinions welcome.

Grüße, Carsten