Observatory
A signal is one record of one item a source’s own publication feed listed: the address the source listed it at, the day the source dated it, and a digest of the listing entry’s own text.
The signal store is a projection kept for reading. The observation series remains the record of what this site has fetched and hashed, and the two are counted apart everywhere below — a signal is not an observation, and every figure names which store it came from.
Current conditions
Seven counts, in the sequence they were filed rather than by size. Ordering seven figures by their own magnitude would put the largest first and present that placement as a finding about the record. Each carries the population it is drawn from and the rule that selects it, and a zero carries the name of the zero it is: a page of bare zeros cannot be told apart from a reader that is broken.
document changes
35 of 132 tracked entries.
a tracked entry whose retained document text changed at least once inside the window, counted through the same per-authority pulse the authority pages render.
Beside it — 131 of 132 carry a document observation at all; a change can only be detected in an entry that has been read twice.
- 27 — over the last 30 days, from
- 35 — over the last 90 days, from
- 35 — over the last 365 days, from
official signals
0 of 0 signal records held.
one record of one item a declared source’s own publication feed listed, held in the signal store.
Which zero — no read has been attempted against any declared source, so this is an unasked question rather than a source that listed nothing.
Beside it — 0 of 2 source(s) carry a successful read. 0 of 2 hold lines this reader could not parse and are excluded from the total, because zero of anything is a failure to read.
- 0 — over the last 30 days, from
- 0 — over the last 90 days, from
- 0 — over the last 365 days, from
consultations
0 of 0 signal records held.
a signal record whose own class field and class state agree on the publisher-stated class “consultation”.
Which zero — no read has been attempted against any declared source, so this is an unasked question rather than a source that listed nothing.
Beside it — counted over all 12 classes the filed order names, so a class with no member keeps its place and its zero is a measurement rather than a key nobody wrote.
future-work items
0 of 0 signal records held.
a signal record whose own class field and class state agree on the publisher-stated class “work item”.
Which zero — no read has been attempted against any declared source, so this is an unasked question rather than a source that listed nothing.
Beside it — counted over all 12 classes the filed order names; there is no enforcement class among them, which the store’s own vocabulary states in its own words.
upcoming events
3 of 17 dated events held.
a calendar entry the publisher itself dated on or after the record day, whatever its date confidence.
Beside it — a date this record holds is the publisher’s, and each carries its own confidence beside it on the calendar; an expected date rendered as a confirmed one is the single most damaging error this record could make.
sources with no change
96 of 132 tracked entries.
a tracked entry this record holds at least one document observation for and in which it has never detected a change, over the whole series rather than a window.
Beside it — 1 of 132 hold no document observation at all, which is this record having measured no change rather than having measured none; and 1 of 132 hold exactly one, which has never been compared with anything, so no change could have been detected in it.
sources with limited text
6 of 132 tracked entries.
a tracked entry from which no prose has been retained — nothing is held to check a quotation against, measured from the store rather than declared from the kind of row. It is a fact about what this record holds and never a claim about how much text the document has.
Beside it — 7 of 132 carry the evidence state “source access limited”, where a publisher’s licence or access policy is what bounds the text; the two counts are different facts and neither is derived from the other.
Activity, in three windows
Every window is measured back from , the same day the board’s check bar ends on and the same day each authority page measures back from. Counts only. Nothing here rates a publisher, compares one with another, or orders a list by volume; where a list has an order, the rule is printed beside it.
Official signals, corpus-wide
| Window | Opens | Records | Which zero |
|---|---|---|---|
| last 30 days | 0 | the signal store holds no record at all, so this window was never asked; which zero the store itself is holding is stated above | |
| last 90 days | 0 | the signal store holds no record at all, so this window was never asked; which zero the store itself is holding is stated above | |
| last 365 days | 0 | the signal store holds no record at all, so this window was never asked; which zero the store itself is holding is stated above |
The store holds 0 record(s) carrying a day this reader could place and 0 carrying none. An undated record is not dropped and is not dated to today; it sits outside every window above, which is why both figures are printed.
By authority
An authority here is the host that served a listing, recorded by the collector and carried through unchanged. A host is not a body and never a person. The population is the hosts the store itself recorded, so a host nothing was served from is not a member that could hold a zero. Alphabetical: ordering by count would rank the hosts against one another, which is a claim this record does not make.
No host is enumerated, because the store holds no record to have been served from one. Which zero the store itself is holding is stated per source below, and it is not one fact.
By topic
A signal has no topic of its own. It acquires one through the tracked entries the publisher’s own declared addresses name, and those entries carry the topic identifiers this record publishes as a convention. So the corpus vocabulary is the ceiling a topic count is drawn from and never the count: the corpus states 24 topic identifier(s) today.
No topic carries a count, because no signal record is attached to a tracked entry. That is not a corpus with no topics — the identifiers are stated above — and it is not a topic nothing was published about. It is the absence of the join that would connect the two.
Document changes, from the observation series
A different store, counted apart. These figures are the per-authority pulses the authority pages render, summed over the 31 authority string(s) the corpus states — not a second count taken here. The sum is the corpus only where authority partitions it, and the module refuses the build where it does not: 0 row(s) carry no authority and belong to no pulse.
| Window | Entries read inside it | Entries whose text changed |
|---|---|---|
| last 30 days | 129 of 132 | 27 of 132 |
| last 90 days | 131 of 132 | 35 of 132 |
| last 365 days | 131 of 132 | 35 of 132 |
Coverage, on both axes, beside the unknown count
Two axes and never one field. Document status is what the publisher asserts, as each entry records it. Evidence state is what kind of evidence this record holds for the entry at all. They answer different questions, and one field answering both is what made a single value stand for seven different situations.
74 of 132 tracked entries carry the status value unknown. Nothing
beside it moves that figure: a coverage count states what this record holds,
and it is not an answer to what a publisher asserts.
Axis one — the status each entry records
| Status | Entries | On the board |
|---|---|---|
draft | 3 of 132 | narrow the board to it |
final | 39 of 132 | narrow the board to it |
in-force | 13 of 132 | narrow the board to it |
superseded | 1 of 132 | narrow the board to it |
unknown | 74 of 132 | narrow the board to it |
withdrawn | 2 of 132 | narrow the board to it |
Axis two — the evidence state
Ordered by the precedence the classifier applies, which is stated here because neither permitted ordering is available: alphabetical scatters the two states meaning a person acted to opposite ends, and descending count would rank the states against each other. The precedence is first-match: an entry sits in the first state on it that the entry earns, and that is not a ranking by strength — the classifier files a row that states no status ahead of one a person checked, because the publisher’s silence is the more informative headline and the verification it also holds is one block down, on the ladder.
| State | Entries | What it says this record holds |
|---|---|---|
| not a document | 30 of 132 | a body, committee or programme — there is no document here to retrieve, quote or date |
| a register | 26 of 132 | retrieved and hashed daily; the tier promises no person-verification |
| unstated by the publisher | 3 of 132 | a surface was read and it states no status term this publisher’s family declares; bounded to the surfaces read |
| verified by a person | 51 of 132 | a person checked this entry against the primary source on a recorded day |
| source access limited | 7 of 132 | the publisher’s licence or its access policy limits what this record takes from the source |
| publisher metadata | 1 of 132 | what is held is metadata the publisher publishes about the document — a catalogue entry or a stage record — not the document |
| observed | 13 of 132 | the document at the address the publisher answered is retrieved and hashed here on a dated series |
| not currently retrievable | 1 of 132 | no retrieval of the document this entry names has produced a digest |
The evidence ladder, rung by rung
Rungs in the order the record acquires them, from what an entry is to what has been checked about it. The order is fixed; a rung’s position says nothing about any entry. Each cell counts entries out of the 132 in the tracker, and each row sums to that figure.
| Rung | ✓ held | △ held in part | ✗ read, and not there | — nothing held |
|---|---|---|---|---|
| identity | 116 | 16 | 0 | 0 |
| publisher | 132 | 0 | 0 | 0 |
| jurisdiction | 132 | 0 | 0 | 0 |
| type | 47 | 0 | 85 | 0 |
| current edition | 32 | 0 | 0 | 100 |
| publication date | 1 | 9 | 0 | 122 |
| lifecycle signal | 19 | 0 | 4 | 109 |
| publisher notices | 0 | 0 | 0 | 132 |
| observation history | 131 | 0 | 1 | 0 |
| text | 91 | 35 | 1 | 5 |
| diff | 37 | 88 | 1 | 6 |
| verification | 58 | 2 | 7 | 65 |
- ✓ — this record holds it, and the sentence beside it names what from
- △ — part of it is held, or it is held under a limit the sentence states
- ✗ — a reading ran and did not find it — an absence this record measured
- — — no reading of this rung is recorded — nobody has looked, which is not the same as looking and finding nothing
The lens, and where it cannot reach
A lens is a saved query carried in the address after the #, and
this site has one grammar for it. The figures on this page are aggregates
over whole populations, and the browser control that applies a lens hides
rows — it recomputes no aggregate. So no lens is applied here and
every figure above is the whole population every time. That is stated rather
than left to be noticed: a full result read as a lens that matched
everything is the quieter half of the way a lens can fail.
- the publisher’s stated status, per tracked entry — expressible as the
statusfacet, and each value above links to the board narrowed to it. - the evidence state, per tracked entry — not on the lens axis — the evidence state is v43’s second axis and the lens axis is the board’s five facets; the two were built for different questions and no row on this site is filed under an evidence-state term. A link that narrowed the board by one would return every row and read as a lens that matched everything.
- the authority a signal was served from — not on the lens axis — an authority here is the host that served a listing, and the lens axis carries no host facet. Each authority has its own page, and that page is where its counts are published.
- a topic identifier — expressible as the
subjectfacet, and each value above links to the board narrowed to it. - a signal class, or a distance from source — not on the lens axis — both are facts about a signal record, and a lens term is a fact about a tracked entry. The signal store publishes its own class pages, where the counts are over the population the term describes.
Sources
2 source(s) are declared by the store contract and are listed whether or not a file exists for them, because a declared source with no file is the never-attempted state and a directory scan cannot see it. Any source found on disk that the contract does not declare is listed too, and marked. Ordering is alphabetical by source id: ordering by count would rank the sources against one another, which is a claim this record does not make.
ema
No attempt is on the record. state/signals/ema.meta.json is absent, so nothing here has read this source yet. That is not
a source that returned nothing; it is a source nothing has asked.
- — — last attempt
- — — last success
- — — robots state, as the collector recorded it
eurlex
No attempt is on the record. state/signals/eurlex.meta.json is absent, so nothing here has read this source yet. That is not
a source that returned nothing; it is a source nothing has asked.
- — — last attempt
- — — last success
- — — robots state, as the collector recorded it
The store, file by file
The directory state/signals/ does not exist. It is stated
here as its own fact rather than left to be inferred from a zero: a
collection that has never run and a reader pointed at the wrong path
produce the same empty page, and nothing else on it tells them
apart.
| Source | File | Lines | Records | Unparseable | No id | count field |
|---|---|---|---|---|---|---|
| ema | absent | — | — | — | — | — |
| eurlex | absent | — | — | — | — | — |
Records is derived from the file: one per line this reader
parsed into a JSON object. The count column is the meta
file’s own field, carried as a claim and never used as a population.
Its writer re-derives it from the whole store each time it writes the meta,
so the two agree unless lines were appended after the last meta was written
— and where they differ, both are shown and neither is reconciled into
the other.
Records
0 signal record(s) are held, read across 2 of 2 source(s). Of those 2 source(s), 0 carry a successful read on the record. What that zero means is stated per source above; it is not one fact.
What this page counts, and what it does not
Every figure on this page names the store it was read from, and figures from different stores are never added together.
- The signal store is the files under
state/signals/: one.jsonland one.meta.jsonper source. The sources listed are the 2 the store contract declares (ema,eurlex) together with 0 found on disk and not declared. It supplies the signal counts, the three windows above them, and every figure in the source blocks and the record table. - The corpus — the tracked entries themselves — supplies the coverage line, both axes, the ladder, and the topic vocabulary a signal count would be drawn against. The observation series supplies the document-change counts and the two conditions derived from them. A signal is not an observation and neither is a tracked entry; a count over one says nothing about the others.
- Every signal count is derived from the
.jsonl, which is the store. The.meta.jsonis the collector’s account of its own last run and is read for one question only: whether a read has ever succeeded. - 0 source(s) are excluded from the total because their file holds lines this reader could not parse. An unmeasured result is not a negative result, and a total that absorbed one would report a failed read as a zero.
- No count is a judgment. A figure here says how many records a store holds and nothing about what any of them concerns, nothing about any publisher, and nothing about whether a number is large. This record reports what documents say and when they changed.
- A title is the source’s own text and is held under
the 14-word ceiling read from
scripts/quote_cap.py. A longer one is withheld with its reason and never trimmed: a truncated quote is a misquote.
The records themselves, by the class the publisher stated and by how far each sits from the document it names, are on the signal store. The same records newest first, filterable, are on activity; what publishers' records place ahead is on the horizon.
Last updated: