How candidates are discovered

The system discovers hypotheses for review, not confirmed transfers. Three categories describe what is inspected; discovery signals describe why it entered the queue. T1–T6 describe the conclusion supported by human review. These are separate classifications.

Candidate discovery flow: preserved sources feed documentary candidates; group screening and shared author addresses feed organizational and human mediation hypotheses; all require independent review.

Open the flowchart at full size. Solid arrows show candidate flow; dashed arrows supply preserved group history/charters or submission and affiliation records. Authorship screening can use documentary endpoints or bypass them through the independent route. Same-ID group history can provide an organizational lead without a supporting document pair.

Three categories, three units of inspection

Category Unit and automatic inputs Discovery signals What remains to verify
Documentary An IRTF-origin / IETF-destination document pair, from preserved references, relationships and affiliation history CIT, CONT, EVOL, FASE Identity, historical affiliations, chronology and actual use of an identifiable contribution
Organizational An RG / WG pair, from saved group screening, supporting documentary candidates, charter excerpts and group history ORG_DOCUMENT_LINKS, ORG_DOCUMENT_CONTINUITY, ORG_CHARTER_REFERENCE, ORG_TYPE_CHANGE, ORG_AGENDA_ADOPTION A process connecting the research agenda to creation or change of IETF activity; a mention is insufficient for T5
Human mediation Documentary-base matches, or independent exploratory document pairs with dated authorship and affiliation observations HUM_SHARED_AUTHOR, HUM_AUTHOR_INTERSECTION Person identity, participation on both sides and a documented mediation role in an independently confirmed T1–T5 event

The categories ask different questions: use of research through documents, organization of an agenda through groups, and a person's possible role in carrying a contribution. The rule that creates a candidate assigns its category. They are not competing labels from which the system chooses a single winner: one underlying case can generate related hypotheses in several categories.

Manual hypotheses can also be registered in each category. They do not fabricate automatic discovery signals. One candidate can have several signals; counting signals is not counting confirmed transfer events.

Documentary relationships

The following mapping uses the preserved Datatracker vocabulary. “Direction” below means the proposed contribution direction; the original factual direction is retained separately.

Family Relationships used Direction and topology
CIT — citation refnorm, refinfo, refold; parsed bibliographic references; refunk as a possible direct lead Cited → citing. Direct citation or one citation connected to definite documentary succession
CONT — continuity replaces, became_rfc, dated historical renames; possibly-replaces as a possible direct lead Earlier → successor. Direct or a directed chain of definite succession
EVOL — technical evolution updates, obs, including normalized inverse forms Earlier → updating/obsoleting document. Direct or a chain of these relationships; identities remain distinct
FASE — institutional phase Explicit forward stream-change events or historical affiliations of the same identity / definite lineage IRTF before IETF. Same-identity direct lead or succession chain; undated same-identity co-observations remain pending

Direct means one admissible connection; chain means two or more. A pair can have both. Two citations in succession do not establish use of the earliest source. Possible references/replacements are not transitive bridges. Publication as an RFC does not by itself establish an IETF destination.

FASE versus human signals: FASE concerns institutional phases of a document identity or definite lineage. HUM_SHARED_AUTHOR and HUM_AUTHOR_INTERSECTION concern an address observed in authorship on two document endpoints. A phase change need not have a shared author, and a shared author does not demonstrate a phase change or mediation. Independent human discovery does not require a documentary relationship.

contains and RFC-index is-also can resolve a citation to an STD/BCP series into a pending CIT chain (RFC member → series → citing document). Historical membership and the component actually used still need verification.

The remaining catalog relations supply context, not independent technical discovery: downref-approval, conflrev, and status processes tops, tois, tobcp, toexp, toinf, tohist. Approval, conflict review and a change of standards status are not proof of a stream transition or technical use. See the complete relationship mapping and theoretical basis.

Organizational discovery

These signals are derived rules, not additional Datatracker document-relationship codes.

Signal Rule and stored topology Main limitation
ORG_DOCUMENT_LINKS An existing documentary candidate connects an observed RG document to a WG document; chain RG → document relationship → WG Current group association is not historical adoption evidence
ORG_DOCUMENT_CONTINUITY The supporting documentary candidate also has documentary_continuity; chain Documentary succession does not prove group succession
ORG_CHARTER_REFERENCE A located RG mention in a preserved WG charter, with verified text/hash and revision observation within the cutoff; direct A name mention does not show agenda transfer
ORG_TYPE_CHANGE Same persistent group ID has verified RG history before WG history, within the cutoff; direct Observations bound the change, not its exact date or cause
ORG_AGENDA_ADOPTION A verified WG charter has an explicit research attribution to the RG and a nearby work mandate, selected by narrow English patterns; direct, possible, pending semantic review Lexical screening does not establish that the passages concern the same work, adoption, or institutional succession; source-work chronology remains for review

The origin must belong to the frozen RG inventory. Documentary support and artifacts must belong to the same run. Charter mentions currently enrich the saved screening universe; this is not an exhaustive independent search across all charters. Groups without documentary links, different-ID succession and BOF/process pathways can remain undiscovered.

ORG_TYPE_CHANGE checks consecutive dated RG then WG observations of the same persistent group ID. Both source records must be verified and fall within the cutoff. It does not detect a different-ID successor. A zero result can reflect the available history and the rule's narrow scope; query completeness does not certify full institutional history. ORG_AGENDA_ADOPTION investigates possible adoption of work and must not be interpreted as formal succession of organizations.

Use Results → Organizational → Evidence between groups for screening context, then inspect the associated candidate and its preserved input snapshot. The current group report and the candidate's historical input can have different dates.

Agenda adoption and charter dates

ORG_AGENDA_ADOPTION enriches the existing RG/WG candidate; it does not duplicate it or assign T5. The rule preserves separate attribution and mandate excerpts. It selects narrow patterns for work developed/defined by an RG or research and development in an RG, plus a nearby WG mandate to update, integrate, adopt, standardize or add work to protocols. Contextual mentions, examination of proposals and cooperation alone do not trigger this signal. The two sentence starts must be within 1,600 characters; proximity does not establish that they concern the same research result. English lexical patterns are partial coverage and can miss valid cases; semantic linkage always requires human review.

Revision observation, publication and charter approval are separate dates. Legacy observed_at remains an observation; it is never substituted for publication or approval. Explicit saved Datatracker new_revision and iesg_approved events can support charter dates. An approval under a draft revision can bind to the next numbered approved revision only when publication shares the exact timestamp. Missing or ambiguous events leave charter chronology pending. Supported charter dates do not establish the date of research adoption or the chronology of the originating work.

Discovery uses saved evidence only. To add explicit events for a named group in bounded, resumable batches:

python web/manage.py collect_charter_chronology --run RUN_UUID --group dtn --budget 8
python web/manage.py discover_organizational --run RUN_UUID

The first command downloads and preserves original event-response bytes and collection metadata. The second enriches existing candidates from saved inputs, preserving reviews and historical input snapshots. Source evidence can be extended on the same frozen run; its cutoff and inventory remain unchanged. Existing runs acquire the new signal only after organizational discovery is rerun with the updated code. Use the Results family filter ORG_AGENDA_ADOPTION to inspect the leads and their evidence gaps.

Human mediation discovery

Both automatic routes use posted draft submissions within the cutoff. They match exact addresses with normalized domains and preserved local-part case. Names alone and inferred aliases are not matched; addresses remain identity hints. RFC authorship is not inherited automatically from draft predecessors. Discovery does not establish a person's role or confirm T6.

Documentary-base shared authorship

  1. Index usable author addresses from the preserved submissions.
  2. For each documentary candidate, find authorship observations on both exact endpoint document names.
  3. An equal address creates HUM_SHARED_AUTHOR, a possible chain: origin document → address hint → destination document, linked to the base candidate. One author list is not duplicated into two phases of the same-document case.
  4. Review identity, affiliations, chronology and actual mediation. This route does not apply the independent route's stricter historical-phase and submission-order screening, so the two queues are not interchangeable.

The base can still be pending when the lead is created. Filter Human mediation by base status to prioritize confirmed T1–T5 cases. A confirmed base does not confirm T6.

Independent author intersections

HUM_AUTHOR_INTERSECTION discovers exploratory human connections without an existing documentary candidate. It first completes the saved-submission authorship index in bounded batches, then uses the latest preserved analytical document and relationship snapshot. Each authorship observation requires an unambiguous historical phase observed by the start of its submission day; the source submission must be strictly earlier than the destination submission. Same-document pairs and unknown/conflicting phases are excluded. Historical group types and continuous affiliations still need verification.

Pairs with an existing documentary candidate or a direct relationship in either direction are separated from this exploratory queue. The remaining exact-address matches produce human-mediation candidates with no base linked and with identity, contribution and mediation explicitly unverified. One candidate represents one address/document-pair lead, not a verified person or a transfer event. Indirect documentary paths are not exhaustively searched.

Use Open exploratory queue in Results, or select Human mediation, family HUM_AUTHOR_INTERSECTION, and base No eligible base linked. The separate discovery action performs no downloads and is not automatically invoked by the group stage’s base-driven pass. Supporting source bytes are verified before candidates are projected. The terminal equivalent is:

python web/manage.py discover_author_intersections --run RUN_UUID --pages 8000

Resume until indexing completes. Repeated discovery preserves IDs, annotations, human-selected base links and reviews. Older leads are retained if later inputs no longer propose them; the coverage status records how many are outside the latest snapshot. Link a documentary or organizational event later from the candidate page with a justification. The link is recorded in the run log and does not itself confirm anything. Confirmation of T6 still requires an independently confirmed T1–T5 base and a located role-bearing passage. Changing a base preserves historical reviews but prevents confirmations about a different base from being applied to the new link.

The original base-driven command remains discover_human_mediation --run RUN_UUID --pages 8000. It has its own coverage status; running either route does not imply that the other has searched its universe. An organizational event can also be linked manually to a human lead. See review dependencies and event association.

What the results do not cover

  • No automatic confirmation of technical, conceptual or empirical use (T2–T4) from a reference alone.
  • No comprehensive analysis of BOF/meeting minutes, slides, mailing lists, person aliases or participation histories.
  • No exhaustive enumeration of all graph paths; discovery selects representative admissible paths.
  • No negative scientific finding from a zero count. Check saved inputs, execution status and source coverage first.

Next: inspect and review a candidate · scientific T1–T6 taxonomy · Results guide.