Relationships and theoretical foundations

This is the relationship mapping implemented by the discovery engine, based on the Datatracker catalog. It is a local screening taxonomy, not an IETF taxonomy of scientific transfer. See discovery for the three categories.

Organizational and human families are application-derived signals rather than Datatracker relationship codes. ORG_AGENDA_ADOPTION selects possible charter attribution/mandate links. HUM_SHARED_AUTHOR uses documentary candidates; HUM_AUTHOR_INTERSECTION can discover independent address-based leads without one. Their rules, dates and limitations are defined in discovery. These signals do not add edges such as cites or replaces to the source catalog.

Why these families remain separate

Basis Methodological consequence
RFC 3967 §1.1: normative and informative references serve different purposes Preserve citation subtype; neither proves which technical object was used
RFC 7841: stream, status, predecessor relationships and series membership are distinct Keep succession, technical evolution, phase and series separate; RFC publication is not automatically an IETF transition
RFC 2026, amended by RFC 6410: standards categories and maturity A status process is not a change of stream; a proposal does not prove its approval/effective date
RFC 8067: normative downref handling Approval is procedural context, not evidence of technical use
RFC 5742: conflict review Institutional interaction is not an adoption decision
W3C PROV: distinguish entities, activities and derivation Preserve source observations separately from derived hypotheses; this is not a claim of full PROV implementation

Exhaustive mapping of the 18 catalog relationships

Factual direction below is Datatracker source → target. Projection is the proposed contribution direction; original fields are never overwritten.

Code Family / layer Projection and permitted interpretation
refnorm CIT, documented Cited → citing; normative reference, actual use still reviewed
refinfo CIT, documented Cited → citing; informative reference
refold CIT, documented Cited → citing; reference without known normative/informative split
refunk CIT, possible Cited → citing; direct review lead only
replaces CONT, documented Replaced → replacement; directed succession
became_rfc CONT, documented Draft → RFC; stream checked independently
possibly-replaces CONT, possible Possibly replaced → possible replacement; direct lead, no lineage union
obs EVOL, documented Obsoleted → obsoleting; distinct documentary identities
updates EVOL, documented Updated → updating; distinct documentary identities
contains Auxiliary: series RFC member → STD/BCP for resolving series citations; no identity merge
downref-approval Auxiliary: downref Preserve approval source → target; contextual evidence, not a substitute for a normative citation
conflrev Auxiliary: conflict review Review process → reviewed document; no technical candidate by itself
tops Auxiliary: status Status-change process → document; proposed move to Proposed Standard
tois Auxiliary: status Status-change process → document; proposed move to Internet Standard
tobcp Auxiliary: status Status-change process → document; proposed move to Best Current Practice
toexp Auxiliary: status Status-change process → document; proposed move to Experimental
toinf Auxiliary: status Status-change process → document; proposed move to Informational
tohist Auxiliary: status Status-change process → document; proposed move to Historic

The six status codes identify the destination status of a process; checking its proposal, decision and effective date is separate work. They must not be interpreted as six kinds of completed transfer. Auxiliary nodes (std, bcp, statchg, conflrev) remain separate from technical draft/RFC endpoints, even when an auxiliary process is associated with the IETF stream.

The RFC index also supplies RFC → STD/BCP is-also membership; its factual direction and code are preserved rather than relabeling it as a Datatracker contains row. Other sources contribute aliases and derived observations: inverse update / obsolescence / replacement forms are normalized; bibliographic section extraction supplies cites_*; a dated rename derived from the same persistent identity supplies historical_rename. Each retains its derivation and original evidence. They are not additional members of the 18-code catalog.

Admissible paths and non-inferences

Let C denote definite documentary succession, B the cited → citing projection, E the old → new update/obsolescence projection, and M membership RFC → STD/BCP. * allows zero or more steps; + requires at least one. A chain needs at least two links overall.

Family Admissible pattern Required limits
CONT C+ Directed succession; no common-ancestor inference
CIT C B C Exactly one citation; provenance identifies which revision/document actually cites which predecessor
EVOL E+ Reachability through updates/obsolescence, not an asserted direct relation or merged identity
CIT via series M B Always interpretation pending until historical composition and pertinent component are verified
FASE Explicit same-identity event, ordered same-identity observations, or C+ with ordered endpoint affiliations Affiliation-derived order uses the earliest known IETF presence; explicit events date a local transition; lineage does not date each link

For example, IRTF A → successor B → citing IETF C → published RFC D is a CIT chain. It does not assert that D retained B's citation or A's technical object. An IRTF RFC belonging to STD S, cited by IETF D, is only a pending component hypothesis: D might use a different RFC from S. Membership observed today does not establish membership at the citation date.

Two successive citations B B do not imply use of the earlier source. Possible references and replacements never serve as transitive bridges. Status, downref approval and conflict review never become generic technical bridges. An approval and a citation between the same pair are parallel evidence, not two steps. Documents sharing a predecessor or series are not thereby successors of one another. No generic graph transitive closure is used.

Time and provenance

Exclude known post-cutoff endpoints/intermediates. Undated relationships remain temporally pending; do not substitute collection time. FASE affiliation ordering uses the earliest known IETF presence, and an explicit event supports that local transition rather than every historical phase. General connectivity does not date each link or prove technical use.

The engine preserves factual endpoints, projected directions, evidence locations and rule identifiers. Deterministic searches select representative paths, not every possible path. Missing historical series membership, status decisions and conflict-review conclusions remain coverage gaps. Original records and human reviews are preserved across reanalysis.