interim-2021-cbor-02
2021-01-27 15:00 - 16:00 UTC

Recordings: https://www.youtube.com/watch?v=6N9mFWsm2z8

Attendees:

• Francesca Palombini, Ericsson
• Christian Amsüss
• Marco Tiloca, RISE
• Peter Yee, AKAYLA
• Carsten Bormann, TZI
• Michael Richardson, Sandelman Software Works Inc
• Barry Leiba, FutureWei
• Ira McDonald, High North/Toyota

Minutes

(Action items noted with “→” symbol and initials)

Intro (CA)

Francesca as next AD. Looking for co-chair.

WG documents status and issues (CA)

• draft-ietf-cbor-tags-oid-04: in shepherd review: waiting on Sean Leonard’s IPR declaration

CBOR implementations survey (MCR)

• would like to go on cbor.io if Carsten agrees, don’t think it should be an RFC
• purpose: several existing implementation and not all of them equivalent.
• example: Rust serde seems good but only good for serializing Rust structures back and forth, no tag support
• describe, not judge, because different impl have different purposes
• possible questions: support for tags? optional? “core” tags only? user defined tags? no API interaction from tags? floating points: supported? mandatory? optional? half vs full/double support? Decoders: pull parsers, “SAX”/event-type, GetNextObject, roll-your-own?

CB: we need to write that draft fast to define different category implementations.
MCR: write a draft, ask people to self identify and tell us where we miss critical bits?
CB: yes

MCR: what to do with implementations where I don’t get responses?
FP: not much more we can do than just list them without features, if they don’t answer.
CA: if there are public issue tracker we can use that
MCR: first question “are you in the mailing list”
FP: On tooling, secretariat could help (already started a survey)
CB: open the wiki feature from the github repo of the draft. start filling it, based on what we know, and wait on them to fill more.

→MCR to start terminology draft.

Packed CBOR (CB)

• update since IETF meeting
• -01 submitted today.
• clear split between the unpacking and the tuble stuff
• added suffix packing
• specify how to apply affix to containers
• defines “basic packed CBOR”
• Suffix packing - allocate more to prefix?
• container affixes - present proposal override; no removal of entries

CA: why would one have suffix override rump rather than having the data in the first place?
CB: rump entirely spelled out that would be weird, but rump could have ref to shared items.
MCR: CBOR protocol executed through a serie of steps.

• prepend in table setup
• bucket idea presented
• next steps: continue discussion, generate -02, complete this without defining all application specific tags, but keeping those already defined in mind.

Ira: Agree with hunch with prefix more important than suffix, and length also longer.

MCR: question about 2.2 and 2.3 tables. Table 1: negative number?
CB: no, N is negative, result is always positive
MCR: does not overlap?
CB: no, alternating
MCR: Ok.

MCR: Table 2 suffix reference. Don’t understand how that works?
CB: Tagging the rump.
MCR: Ok, so you are tagging the rest. Now I get it. Text need to be clearer.
FP: review would be great.
MCR: Yes.

MCR: ??
CB: for the first 16 values you have a single value encoding. Next you have 1 byte.

→CB: more examples
→MCR: review on mailing list?

CB: have the next version of the draft pretty quickly so please keep comments coming.

→CB to clean his implementation and share it.

Discriminated unions

• came as a need from implementers from language environment side
• tags 27, 1001 already defined
• presenting proposal
• not yet submitted draft, feel free to have a look and provide feedback

CA: Anybody read the pre-draft?
CB: First time shown.

AOB

WG: Thank you Barry!