CBOR at IETF111

Agenda

(The 2' slots for some documents are primarily to state document status / aliveness, and/or get room input on further directions, based on the standing assumptions that People Have Read The Drafts.)

Minutes

Minute takers: Marco Tiloca

Christian Amsüss: introduction and chairs' slides

RFC 9090

RFC9090. Was cbor-tags-oid.

Christian Amsüss: Released 2 weeks ago

cddl-control

cddl-control Processing WGLC and shepherd.

Carsten Bormann: Completed WGLC. Still to process Christian's comments towards v -04, then ready for IESG submission.

file-magic

file-magic: WGLC will need more feedback to complete.

Carsten Bormann: Nice document but no feedback from WGLC. Please, have a look at it. Carsten Bormann: Do we want to do the other tag registration? Content-format tag in the appendix? We need feedback.

John Preuß Mattsson on jabber volunteering to review.

network-address

network-addresses: WGLC ending here, shepherd writeup ready.

Carsten Bormann: Good discussion ongoing, converged on format of addresses and prefixes. Carsten Bormann: Do we want to address "zone-ids"? That opens for possible mistakes though. Carsten Bormann: A new version should be soon ready to submit. Will move forward together with the file-magic document.

time-tag

time-tag. Just adopted recently.

Carsten Bormann: Recently adopted, stable for a while. Anything we want to add, or so stable and good already for defining more properties? Carsten Bormann: Not in a rush, tags are registered and used. We need reviews.

Bron Gondwana: I will review, also to ensure cross-compatibility

packed

packed

Carsten Bormann: Around for a while, waiting for more independent implementations. Needed to fix the "table building" issue. Also good for trying over different use cases - These include C509 certificates https://datatracker.ietf.org/doc/draft-ietf-cose-cbor-encoded-cert/ Please implement and try. Carsten Bormann: Possible split between tags and table setup. Maybe even without a table setup in the first document as all, or just simple basic (like in current document).

Brendan Moran: From "what SUIT would have done", we'd need nested dicts, and not just "one preshared and one received", but 2 or more in a single manifest envelope. SUIT has few pieces put into a container; one is signed and others authenticated. Removable pieces would make use of table in the signed part. Other parts might have other separate metadata, that gets discarded. Tree of tables.

Christian Amsüss: You mean that the definition of the tree would be used outside of the tree as well? Brendan Moran: Root of tree is shared table, but each of leaves have own table, because they're detachable. Never used separate from root.

Chris Lemmons: Potential use case involving JWTs and CWTs. Don't think you can pack CWTs. Carsten Bormann: Any document you can point to? Chris Lemmons: No and it's about outsider customers Carsten Bormann: We'd need to well understand the use case anyway Chris Lemmons: Will try to bring more info

Bron Gondwana (on chat): Feels like XML namespaces. Carsten Bormann: We were pushed forward by the CoRAL document in CoRE https://datatracker.ietf.org/doc/draft-ietf-core-coral/ which does something similar than xmlns / CURIEs. Nothing really new.

Chris Lemmons: Might be used also for efficient document transfer. A client asks for XML/JSON/CBOR. Current no way to signal that I want packed CBOR. Something like cbor+packed Carsten Bormann: Can be one more structured syntax suffix, though they don't nest but it should be ok Christian Amsüss (from the floor): isn't it like for any CBOR extension, not only packed? That shape can mandate pack CBOR or not. Not sure a structure suffix helps here. Chris Lemmons: Could be infinitely detailled -- could describe all extensions and all types etc, and that's crazy. But packed CBOR feels special to me. It's a big component. And complexity is different ways of encoding it, how smart are your table formats. On parsing side, it's more straightforward, you follow rules. Feels different to me, but maybe not.

Carsten Bormann: You can combine cbor-packed with other CBOR specs, so it's orthogonal. A bit about content-coding but not exactly. But kind of orthogonal. See certain value in this as structured suffix, but also the explosion this can cause. Carsten Bormann: Need to make this an acceptable feature is a good requirement and feedback.

notable-tags

notable-tags. Including discussion on "how to maintain usability of tags registry". Carsten Bormann, 10'

Carsten Bormann: Individual document, more a skeleton now than an actual draft. Carsten Bormann: Provide an overview of most important tags; it can also be more opinionated than registries, to "recommend" what best to use among similar tags (see e.g. tag 260) Carsten Bormann: Expected to grow over time and to take a couple of years to be in a useful state; no rush but need to work on it Christian Amsüss: Checked with IANA that it is fine to refer to this document in the registries, once a WG document Ira McDonald on Jabber: Maybe not publish notable, but use wiki? Carsten Bormann: Might also be not published and rather a wiki, but a document is better as coming with additional control

cddl-freezer

cddl-freezer. Priorizing frozen goods. Carsten Bormann, 10'

Carsten Bormann: Generated before publishing the CDDL spec. Collection of good ideas put in the freezer for the next version. Lot done as to control operators. Need to actually touch the language again? Carsten Bormann: Possible actions: i) more control operators but they might remain in the freezer; ii) actual changes to the language like cuts, plus errata and clarifications, namespaces and cross-universe references; iii) more CDDL independent points e.g. alternative representations (eg. a canonical AST for tools to work with; freezer has workable AST right now.)

(p11) Namespaces

(p12) Use of updated CDDL syntax in existing RFCs

Brendan Moran: Plans to handle specs that kind of define CDDL but not actually, eg. COSE, where you'd need to run a script. Carsten Bormann: That works Brendan Moran: In Python it doesn't. Carsten Bormann: Currently using Perl .... but yeah.

Carsten Bormann: So idea is to have space where these exported CDDLs can be used from. Brendan Moran: Strong preference for using API from IANA Carsten Bormann: Link? Brendan Moran: Will provide

Carsten Bormann: We may want to include even other documents eg. European Digital Green Card.

(p13) Definition of "import" and "export" operations. Could use XMLRFC file name of the examples to decide whether it's exported or not.

map-like

map-like. Carsten Bormann, 2'

Carsten Bormann: In a solid shape now; perhaps address also lists and sets? Brendan Moran: I case about one use case, namely the list of pairs, just as automated items in the list Carsten Bormann: Already in the document Brendan Moran: That's the one I really care about

Christian Amsüss: Want to check who has checked the document so far, among about 30 participants 5 "Raise hand" (read a version) 2 "Don't raise hand" (I don't care) Christian Amsüss: Appreciate more opinions on the mailing list; looks like low-profile now

Carsten Bormann: Not done before to avoid reinventing things Brendan Moran: It's not as not easy as it sounds; I needed this, you need this if you care about code generation Carsten Bormann: Agree with that

(p16) "Records" Carsten Bormann: Two items discussed on Github, perhaps worth considering under cbor-packed, while the rest can be handled with semantic tags or something. Would like to have feedback on the way forward.

Christian Amsüss: New question for "show hands". Should this be pulled in cbor-packed? 3 "Raise hand" (prefer to pull-in) 3 "Don't raise hand" (keep separate) Christian Amsüss: Need more discussion on the mailing list

AOB

Christian Amsüss: Plan to have interim meetings biweekly, alternating with the CoRE interim meetings, every other Wednesdays at 16:00 CEST. Roughly to resume in September.