CBOR WG Conference Call
Wednesday, Dec 19 2018, 18:00 - 19:00 UTC
Chairs: Barry Leiba, Francesca Palombini

Present:
Barry Leiba
Carsten Bormann
Jeffrey Yasskin
Paul Hoffman
Sebastian Hans

Regrets:
Francesca Palombini
Laurence Lundblade

CDDL
- Edit has addressed all IESG COMMENTS in his version
- One DISCUSS from Eric still outstanding
- Paraphrase PEG so no normative reference is needed
- Need to confirm that this solves everything

CBORbis
- About 10 issues and 3 pull requests
- Thanks for Jeffrey for his reviews and doing the pull requests
- Need to talk about well-formed and valid
- Current text explains the difference between the two
- Text is changing for rules on validity of tags
- New Req’ts proposed:
  1. An encoder MUST produce only valid items.
  2. A decoder MUST stop and return an error with no data when it
encounters an ill-formed item.
- Req’t 2 is straightforward, but req’t 1 is problematic
…discussion…
- Jeffrey to rephrase req’t 1 to make it less problematic
- Editorial: do we need “ill-formed” and “invalid” as formal terms for
completeness?
- Consensus is no, use “not valid” and “not well-formed”
- Editorial: we have to be more specific about what we call
components, and avoid terms such as “parser” except in only the
general sense
- A number of other editorial improvements…
- Technical: base64 padding issue; boundaries between basic and extended types
- Outside the draft, need to work on test vectors, especially on
failing test vectors
- Should we include some of those in the draft as well?
- Put technical issues into individual threads on the mailing list
- Test vectors may be good fodder for hackathon, with preparation on
the list now
…discussion of how to get people involved with this…
- Carsten will use datatracker to check references to CBOR to get a
start on whom to contact
- Jeffrey brings up the issue os “ill-formed” vs “fatally ill-formed” (pr #41)
- Duplicate map keys, for example… discussion of that.  Maybe call
that “not valid”, rather than “not well-formed”?
- Jeffery will start a thread on the mailing list

2 Jan teleconference: cancel because little will happen over the holiday weeks.