Skip to content

docs: sweep to the Anaconda 2026.06 baseline and measured container sizes #106

Description

@mmcky

Docs are a release behind the containers and three of their claims about the images are wrong (2026-07 maintenance audit). The CHANGELOG half is closed and PLAN items 3/4/13 struck; no container-facing doc text has changed, so §1, §3 and §4 stand as filed.

1. Eight locations state the containers pin Anaconda 2025.12; both ship anaconda=2026.06

Ground truth: containers/quantecon/environment.yml:11, migrated in #95. The lean image carries the matching anaconda=2026.06 linux-64 py313 (MKL) pin set (comment at containers/quantecon-build/environment.yml:12-19, pins from :20) — same baseline.

Stale: containers/README.md:18, containers/quantecon/README.md:15 and :107, containers/quantecon-build/README.md:26 and :82, docs/ARCHITECTURE.md:12, docs/CONTAINER-GUIDE.md:9.

PLAN.md is now clear: the dependency policy was corrected in #139 and PLAN.md:14 in #127, leaving only the deliberate narration at PLAN.md:68 of the date-pin misconception that produced #86/#87.

CHANGELOG excluded — its 2025.12 mentions at CHANGELOG.md:254, :260 and :553 are correct historical record, and :260-261 retro-annotates the 2026.06 move.

2. CHANGELOG — closed

The #28 entry was rewritten in past tense and retro-annotated at CHANGELOG.md:254-261; #95 landed at CHANGELOG.md:203-207, #90 at :208-209, #94 at :210 — all in v0.9.0 (#113). #91 (rewrote PLAN.md and README.md) and #93 (added PROJECT-OPTIMIZE-PREVIEWS.md plus a README index link) were dropped as docs-only planning edits; the repo has not changelogged the equivalent since (#130 and #133 are unlogged PLAN.md edits).

3. Six docs still advertise "~8GB / ~3GB" and "~60% smaller"

Measured — already correct at README.md:147-148, docs/ARCHITECTURE.md:15 and :53, docs/CONTAINER-GUIDE.md:9-10, fixed there by the v0.8.0 sweep (CHANGELOG.md:337-339) and nowhere else — is 8.3 GB / 7.1 GB on disk (~3.2 / ~2.9 GB compressed): a ~14% saving, not 60%.

Nine stale sites: docs/QUICK-REFERENCE.md:22-23, setup-environment/README.md:34-35, containers/README.md:9-10 and :35 (the one surviving "~60% smaller"), containers/VALIDATION.md:17-18, containers/quantecon-build/README.md:18 and :79, .github/copilot-instructions.md:33-34 and :134.

Do not copy the pair forward unchanged: v0.11.0 changed the measurement basis (#108) — the size job at .github/workflows/test-container.yml:97-101 now reads the registry manifest and reports compressed layer bytes, a smaller number not comparable with the old docker images series (CHANGELOG.md:83-84; same warning at .github/workflows/test-container.yml:73-76).

4. "Minimal TexLive" is false

Diffing the apt lists at containers/quantecon/Dockerfile:8-44 against containers/quantecon-build/Dockerfile:9-45 gives exactly two differences: the lean image drops texlive-luatex (containers/quantecon/Dockerfile:15) and adds graphviz (containers/quantecon-build/Dockerfile:32); texlive-latex-extra and texlive-fonts-extra are in both. LaTeX dominates both, hence the small size gap; the claim is the stated reason for the lean image and the basis of every "use quantecon-build for CI" recommendation. Stale at containers/README.md:33, containers/quantecon-build/README.md:15, containers/VALIDATION.md:18 and .github/copilot-instructions.md:34.

5. PLAN.md and CONTRIBUTING.md

Items 3/4/13 are struck with closure notes at PLAN.md:68, :51 and :70 (#130). Left: the "Open issues disposition" table at PLAN.md:108-117 carries rows only for #83, #123, #14, #29, #30, #18, #27 and #2 — nothing opened since the July 2026 review. The doc-update matrix at CONTRIBUTING.md:129-134 lists only the action README.md, docs/QUICK-REFERENCE.md, docs/MIGRATION-GUIDE.md and docs/FUTURE-DEVELOPMENT.md — no PLAN.md row, which lets the backlog drift between audits.

Checklist

Sequencing

docs/QUICK-REFERENCE.md and setup-environment/README.md are also edited by #109 (#99 folded into it), and PLAN item 10 at PLAN.md:80 carries "Propagate corrected container-size figures everywhere" over the same files — land them as one docs PR, not three passes.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    Type

    No type

    Projects

    No projects

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions