Working resources: databases, research notes and the order of references

A working resource is anything that lets a claim be found twice. This page sets out how a documentary desk moves a fact from its first note to the version that can be cited: which research databases propose candidates, which search tools widen the field, how a note is taken so it still means something a year later, and how references are ordered so that a reader can retrace the whole path without asking the desk.

The habit is not specific to discography. A Serbian publication about research methods keeps a comparable shelf, where databases, note organisation and the preparation of talks are treated as one continuous craft, and a Serbian guide to research work describes the same path from a first jotted question to a presentable, checkable version. The language differs; the discipline of a findable trail does not.

What a database is for, and what it is not

A database proposes candidates; it does not settle them. When the ledger consults a structured music record, the row that comes back is a hypothesis about identity: this title, this artist, this catalogue number, this year. The hypothesis becomes usable only when a second, independent route agrees with it. A row that stands alone stays in the working file with a flag, and it stays there until something outside the same database family confirms it.

This is why the desk treats database output as lead material rather than as text to copy. A catalogue number seen in a community record is worth writing down with its provenance attached; it earns a place on a public page when the same number survives contact with an artist-controlled page, an institutional biography or a dated archive capture. The rule feels slow. It is also the reason a reader can trust what the ledger prints, and the reason a correction, when one is needed, has somewhere precise to land.

The first note decides everything after

A note taken badly cannot be repaired later by a good filing system. The desk keeps a small rule for the first contact with any source: record what was seen, where it was seen, when it was seen, and whether the page stated the fact or merely implied it. Verbatim wording and the reader’s interpretation of it are two different entries, and the note keeps them apart from the beginning.

For capture itself the desk uses the Zotero reference manager, a project whose own documentation draws the same boundary: the item record, the snapshot of the page as it looked that day, and the researcher’s annotation are three separate objects that happen to travel together. A working file that confuses them cannot answer the only question that matters later, which is whether the source actually said the thing or whether the reader supplied it. The snapshot also survives the day a page changes or disappears, which on a small-label trail is not a rare event.

Organising references so they can be found twice

A reference is not finished when it is saved; it is finished when it can be cited. Cited means the entry names its edition or capture date, its relationship to neighbouring sources and the field of the record it supports. The working file keeps references in the order they were confirmed rather than the order they were found, because the confirmation date is what a correction will later argue with.

The naming convention matters more than the software. A reference called “discogs maybe 2003” decays within a month; a reference called by artist, title and retrieval date still reads correctly years later. Each entry in the file answers three questions without being reopened: what it is, where it came from and which claim it was collected for. When a page is published, the cited version of each source is the one that appears in the source drawer, and the working copy keeps the rest of the trail for the next revision.

Search tools and their limits

Search widens the field and, left alone, narrows it badly. A query built from a quoted title, a catalogue string or an old route fragment finds different material than a query built from a description, and the desk uses both deliberately. When live search fails, the path continues through archived routes and library catalogues, which have their own biases: they preserve what was capturable, not what was most true.

Every tool also has a silent zone. Community databases are strong on popular releases and thin on tiny pressings; institutional catalogues are strong on dates and thin on session detail; artist pages are strong on credit wording and weak on disconfirming evidence. The working file records which silent zone each source occupies, so that an absence can be read correctly later: a missing row may mean the release was never catalogued, or may only mean the tool never looked there.

The same discipline inside this ledger

The result is visible across the site rather than described only here. The boundary rules are written out in the verification guide, the kinds of sources and what each can prove are filed on the source shelf, and the applied outcome — a title, its credited roles and the trail behind them — is what fills the Session Ledger. A working file that no one can retrace is a diary; the point of the resource shelf is that the route stays public.

None of this is large. It is a handful of habits applied every time: capture with provenance, separate statement from reading, keep references in the order they were confirmed, and publish only the cited version. A reader who can retrace a claim will forgive a gap in the record, because the gap is visible too. That, more than any single database, is what the working resources are for.

From the working file to the public page

The last step is the transfer, and it is where a desk shows its discipline most plainly. Nothing moves from the working file to a public page as a loose fact: it moves as a claim attached to the reference that supports it, in the cited version, with the retrieval date that lets the next reader check whether the source still says it. What does not fit that shape waits in the file, flagged, until the missing confirmation arrives or the claim is honestly dropped.

This is also the step where the order of references stops being private housekeeping. A public source drawer presents them in the order a checker will want them — the strongest for identity first, the corroborating trail after — and leaves the speculative ones out entirely rather than dressing them up. The reader is not shown everything the desk collected; the reader is shown the version that can be defended. The rest of the trail stays in the working file, where it remains useful precisely because it was never published half-finished.