The ecosystem dashboard, and rankings that say what they rank on

[HarryCordewener]

Spec §9's ecosystem dashboard and rankings, both in §14's v1 boundary and neither previously
existing.

Stacks on #11 (chore/park-the-importer), which is main plus #9 and #11. Those two rewrote how
the site gets its data, and building on main would have meant writing against the fixture and stale
wording. Review after #11 lands, or read the diff against that branch.

What is here

Two pages, /ecosystem and /rankings, both static SSR with no script, both surviving ?plain=1
whole. Aggregates live in NpgsqlGameQueries behind IGameQueries; MUI.Catalog references nothing
new.

Shares, never totals. §15.7 withholds the absolute "how many people play MU*" figure because a
ratio over the games we reached survives the ones we cannot reach and the ones nobody claimed, and a
headcount survives neither. EcosystemDashboard has nowhere to put one — it has no access to
presence at all — and NoAbsolutePlayerFigureIsEmittedByEitherSurface sums the fixture's measured
counts and requires the number never to appear on either surface.

A share carries its denominator as a field. MeasuredShare(Label, Count, Denominator), with
Fraction computable from nothing else, so a percentage cannot travel to a renderer without the set
it is a fraction of. Protocols carry two denominators, because they are two sets: games whose
handshake we completed, and games whose MSSP report we hold.

The protocol denominator is read off availability, not off capability rows. A reachable
interval is a session of ours that completed, which is exactly what a measured capability is a
capability of. Counting "games that have capability rows" would define the denominator out of the
numerator — a game whose handshake completed and offered nothing measurable would drop out of the
bottom of every fraction and quietly raise every share on the page.

What the stored data could not honestly support, and what the page says instead

• No adoption curve. game_field holds a current value with transitions beside it in
  field_change, so a curve has to be reconstructed from transitions and there are not yet enough to
  plot. The available alternative — plotting each observation's first_seen_at — would draw a
  confident rising line measuring our crawler reaching more games and nothing whatever about anybody
  adopting anything. The page says so and prints the transition count, which is the number that says
  when the curve becomes real.
• TLS is unmeasured and reads as unmeasured, not as 0%. The probe dials plain telnet, TLS is not
  a telnet option, and nothing writes a tls endpoint — so no capability.tls.measured row can
  exist today. ProtocolAdoption.Offered is nullable for exactly this, and the null is derived from
  the tally rather than compiled in, so the column starts reporting a share on its own the day the
  first measurement lands. TLS still shows a declared share, over its own denominator.
• The measured column is a floor, and is labelled one. FieldObservations writes a capability
  when it observes one and otherwise writes nothing, because the crawler requests MSSP alone and
  declines MCCP outright. MSSP is the one protocol whose silence is an answer — we ask by name — so
  it is the only one whose absence is counted as a no. Everything else lands in neither offered
  nor asked.

Rankings

Two tables, each stating its basis on the page: busiest by the median of measured concurrent
counts over a named week among games that produced at least 24 counted samples (percentile_disc, so
the median is a number some probe actually read), and the current unbroken reachable spell, which
carries the date it began because a spell cannot be longer than we have been watching. Archived games
are out of both and of nothing else. An uncountable probe is not a zero in either; a measured zero is
a count and ranks. There is no voting affordance and the page says there never will be.

Tests

670 across the five suites, all green and warning-free. Worth naming:

• NoAbsolutePlayerFigureIsEmittedByEitherSurface
• EveryPercentageOnTheDashboardArrivesWithItsDenominator
• AGameWeHaveNeverHandshakedIsOutOfTheDenominatorRatherThanInItAsANo
• AHandshakeThatOfferedNothingIsStillInTheDenominator
• AProtocolNothingHasOfferedIsUnmeasuredRatherThanNoughtPerCent
• OnlyAProtocolWeAskedForCountsSilenceAsANo
• AnUncountableProbeIsNotAZeroInARanking

Two things outside the feature

• The fixture's demo week is scaled to each game's own count. It peaked at 17 for every game
  regardless of the count that game reports, which made the demo ranking five ties.
• The test container's client limit is lifted to 500. A database per test met Postgres's default 100
  on the run that took the suite past 130 tests, failing whichever tests happened to be starting and
  reading as a flake in twenty unrelated places.

Deliberately not here

No /api/ecosystem — §10 is a separate surface and this adds no endpoint. No recency window on the
protocol denominator: a listed game we handshaked six weeks ago is in it, and the page states how old
the stalest handshake in the picture is rather than silently trimming the set.

🤖 Generated with Claude Code

https://claude.ai/code/session_01NrGKmKcRCGktyhRTFbQDMk

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: ASSERTIVE

Plan: Pro

Run ID: ba893d4b-3eac-43dc-b538-9f7401b72a05

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

_{Comment @coderabbitai help to get the list of available commands.}