still seedinglxnaydesign.net

Long-form histories of Linux distributions — how they started, how they ended, and what still runs. Written from release notes, mailing lists and archives, not from memory.

Page
Standing policy and log
Published
Verified
Entries
6

Corrections

This page went up before the first article did, because a corrections page created after the first embarrassing error is a press release, and a corrections page that predates the first article is a policy. The log is at the bottom.

The policy

Everything on this site is written from records, and records get misread. When that happens here, four things happen in this order:

  1. The page is fixed.
  2. A dated note is appended to the bottom of the page saying what it used to say and what changed.
  3. An entry is added to the log below.
  4. Whoever reported it is credited by name, unless they asked not to be.

What does not happen: silent edits to a factual claim, deleting a page to make an error disappear, or changing a published date to make a correction look like an original. If a claim was wrong for eight months, the record says it was wrong for eight months.

How errors are classified

Typo
Spelling, grammar, a broken link, a formatting fault. Fixed on sight, not logged. Nothing a reader believed was affected.
Clarification
What was written was defensible but was read the wrong way by more than one person. The wording changes; the claim does not. Logged.
Correction
A factual claim was wrong — a date, a version, a name, a base, a number, an attribution. Fixed, noted on the page, logged, and if the claim also lived in a Register row, that row is fixed the same day.
Retraction
The central claim of a piece does not hold. The piece stays online, with the retraction stated at the top of it above everything else, and the reasoning that failed left visible. Nothing is quietly unpublished.

Staleness is not an error, but it is a debt

Half the claims on a site like this have a shelf life. A distribution that was shipping in August is not necessarily shipping in March. Every article carries a last verified date in the rail and a Standing block that separates what is still true from what has ended and what is now actively wrong; the Register carries a verified date in every row.

A page that has gone stale has not been corrected, it has been neglected — so re-verification passes are logged here too, marked as such, so you can see when something was last actually looked at rather than last touched.

The log

Newest first, each with its date, the page affected, its class, and what changed.

Your distro just died: the first thirty days · Correction. The paragraph introducing the piece promised “the in-place conversion worked end to end”. Ten lines above it the same page’s method block says no machine was migrated to produce this page, and the sources note at the foot says nothing on it is a report of first-hand use. “Worked end to end” was meant in the sense of worked through — the conversion is set out as seven stages with a rollback for each — but on first reading, in the sentence that tells a reader what they are about to get, it says the author ran the migration and it succeeded. That is a first-person experience claim for something nobody did, which is the one sentence type this site cannot publish and the class of claim 16 CFR 465.2(a) exists to reach. It now reads “the in-place conversion worked through stage by stage from the documentation”. Found in a compliance audit of the whole site rather than reported by a reader; it is the same fault, on a different page, as the “about twenty minutes” timing claim corrected on Booting a 2006 live DVD earlier the same day. Class: correction.
Six pages · Correction. A whole-site read of all ten articles, one after another, looking for the errors that only show up when the pieces are read against each other rather than one at a time. Four of the six are places where two of my own pages disagreed. Gentoo ships 89 GB of binaries now gave Sabayon’s last release as “19.03, on 31 March 2019”, in the body and again in Standing. That is the announcement date carried by Wikipedia and DistroWatch, and this site spent a section of the release history establishing that it is the third of three dates: RELEASE_DATE_19.03 on ibiblio says the images were built on 22 March and their file dates put them on the mirrors on 27 March. Both mentions now carry the dated facts, and the Wikipedia entry in that page’s source list says plainly which of its dates I do not use and why. Your distro just died stamped the same beat 2019-03-31 and now stamps it 2019-03-27. What happened to Sabayon Linux still listed, as an open question, whether the October 2021 and February 2022 MocaccinoOS alpha images were built against the Funtoo tree — a cell the successor audit closed the same day, from the build definitions at the two commits current when those images were tagged. And the comparison table in The five candidates called MocaccinoOS’s cadence “monthly, unbroken” two paragraphs above its own body text noting the missing January 2026 release; the cell now names the gap. Two more are internal to one page each. Funtoo announced its own death and then took it back ended a paragraph “Funtoo’s hobby mode ran five and a half” with no unit, directly after a sentence measuring TenFourFox in years — the figure is five and a half months, and the sentence now says so and gives both dates; its hero caption said five. And the opening paragraph of Booting a 2006 live DVD on 2026 hardware promised the exercise “takes about twenty minutes”, a timing claim on a page whose method block states that no image was booted to write it — the same fault two other sentences on that page were corrected for on the day it published. It now describes the shape of the route and says nothing was timed. All six found in my own read-through, none reported by a reader. Class: correction.
The board: what a one-man distro’s forum was actually reporting · Structural. Not an error, but the one article on the site that ended by telling readers to go and look at the front page. Every other piece here closes with a hand-picked Previous / Next / Also block whose entries deep-link into the section of the neighbouring article that actually continues the thought. This one now does too: the release ladder that produced the reports it takes apart, the shutdown timeline it stops short of, and the four failure clusters it counts, met from the other end in the piece on booting one of these discs today. Class: not a correction — recorded here because the standard is meant to be uniform and was not.
MocaccinoOS: the successor, or the estate? · Correction. Six claims, found in an adversarial re-check of my own page rather than reported by a reader, and all listed in a dated note at the bottom of the article. The two that changed what a reader would believe: the page called sabayon-init and sabayon-live-boot “the live-media boot glue — the code that gets you from a booted ISO to a desktop, which is the one piece of a live distribution nobody else writes for you”, when the published specs show both are Ivan Davidov’s Minimal Linux Live at a pinned commit — one differing from its lml- twin by four sed renames, the other by a greeting, a motd and a default root password — so the surviving Sabayon code is one dracut wrapper, not three packages; and the authorship breakdown of Luet’s fifty most recent commits gave figures summing to 44 and understated the project lead’s own share by two, which is now printed as a record from the GitHub API. Also corrected: Luet’s start date, given as 5 June 2019, which is when its GitHub repository was created — the first commit in its history is dated 21 September 2018, from a @gentoo.org address, six months before Sabayon’s last release; a claim that the project’s GitHub organisation calls MocaccinoOS Gentoo-based, when the organisation has no description and the documentation is where the sentence actually lives; two record blocks that attributed build steps to definition.yaml when those lines live in each package’s build.yaml; the Funtoo collaboration’s length, given as thirteen months when it was twelve months and sixteen days; Entropy’s last push, ten weeks before the rebrand rather than two months; the signature block’s span, three years and eleven months rather than four years; and the 349 and 356 package figures, described as repository “badges” when they are the lengths of the published index listings. In the same pass one not established item was answered — the October 2021 and February 2022 alpha images were built from a Gentoo stage3 and a pinned gentoo/gentoo tree, read from the build definitions at the commits current when each image was tagged — and the 349 figure gained a category tally showing 178 of the 349 entries are firmware. Those two are Standing updates rather than corrections and are recorded as such on the page.
Booting a 2006 live DVD on 2026 hardware · Correction. Eight items, found in an adversarial re-check of my own page rather than reported by a reader, and all listed in a dated note at the bottom of the article. The two that changed what a reader would believe: the page said the per-release kernel .config files this domain published “appear not to survive anywhere”, and filed the question of whether a 2006 disc can see an AHCI SATA controller as not established on that basis. All five configs fetch from the Internet Archive intact, and CONFIG_SCSI_SATA_AHCI=y is in every one of them, across both architectures and three kernel versions — a hedge became an answer, and the article gained a section built on the files themselves. Third, the archive.org Gentoo 2006.0 item was described as having “no checksum of any kind”; the Archive publishes an MD5 and a SHA-1 for every file it stores, and the honest claim — no checksum from the publisher — is narrower and different. Also corrected: the assertion that every boot option in the failure table is documented in Gentoo’s handbook, when cdroot lives on the genkernel page and the handbook spells the USB pause option slowusb rather than this project’s 2006 doslowusb; an inexact scandelay quotation; a file size given in mebibytes as though it were megabytes, and an uploader described as private; an assumption that this domain published a torrent for every 2006 release, replaced by the one torrent that survives, parsed here, with the 1,386 piece hashes it still carries for a release whose .md5 did not survive; a Firefox version stated for a disc no manifest covers; and a Phoronix publication date given as 2017-11-20 when it is 2017-11-18. Two sentences predicting what the reader would experience, on a page whose method states that nothing was booted, now argue from the build’s own numbers and say that nothing was measured.
Gentoo ships 89 GB of binaries now · Correction. Three claims, found in an adversarial re-check of my own page rather than reported by a reader. The page said the project’s 89 GB figure was “the aggregate” of the binary host’s amd64 indexes, and that the generic index was “by far the largest” of them. Both fail against the host’s own files: the eight amd64 indexes sum to 172.7 GB, and x86-64-v3 comes within a few per cent of the generic index rather than being a fraction of it. The article now publishes every index’s instance count and size, and what the January figure actually counted has moved into its Standing block, unanswered rather than assumed. Third, the note that Entropy’s development had NLnet funding sat under a Wikipedia citation, and the Wikipedia article does not mention NLnet anywhere; it is now sourced to NLnet’s own project register, which dates the entry to 2009 and states the design target the note was reaching for. A dated note is on the page, with four clarifications beside it: the 2023-to-2026 signature history, the reason official-binhost users needed no action in May 2026, the completeness of the equo translation table (twenty-one commands under a heading promising all thirty-three), and a Gentoo Foundation income figure quoted without the SPI income or the bank balance printed beside it in the same document.
The five candidates for “Sabayon in 2026” · Correction. Four claims, found in an adversarial re-read of my own page and all listed in a dated note at the bottom of the article. Appendix A described Funtoo as “alive but explicitly promising nothing”; it ended for the second time on 5 September 2025 and its tree froze on 29 January 2025, which is exactly the error this site’s own Funtoo piece files under actively wrong. The table and body gave MocaccinoOS 26.08 desktop versions — “GNOME 49.7, Xfce 4.20, MATE 1.28” — that no source states; the release note names only Plasma 6.6.6 and COSMIC 1.5.0, and the documentation versions none of its twelve environments. The page credited Luet with “snapshot rollback and snapshot pinning through its Vajo front end” in 26.08, where the release note says only that Vajo development continued. And it said Calculate Linux had been shipping “a year longer than Sabayon managed in total”, which is arithmetic: it is about six years longer. In the same pass, two cells moved from not established to answered — MocaccinoOS’s base (Gentoo, from the build definitions) and its maintainer count (four) — which are Standing updates rather than corrections and are recorded as such on the page. The Redcore package count and the 2601 desktop versions were also re-sourced and are now dated and attributed rather than stated flat.
RR4 → RR64 → Sabayon: the complete release history · Correction. Six items, all found in an adversarial re-check of my own page rather than reported by a reader, and all listed in a dated note at the bottom of the article. The two that changed what a reader would believe: the ladder claimed no dated release record survived below 2.60, when in fact the author announced 2.40, RR64 2.50, 2.60, 2.60.3 and the 2.60.4 patch with dates in the Gentoo Forums thread, which is still online — six dated 2005 rows now replace two undated ones, and the Register row was corrected the same day; and the page said only four 2005–2006 checksums could be recovered, when six come back intact. The extra checksum for the RR64 3.0RC1 miniEdition also evidences a release the ladder had missed entirely. Three artifact claims that said “survives” where only a filename is indexed, one mis-cited archive capture date, and one release-range start date were corrected at the same time.
The five candidates for “Sabayon in 2026” · Correction. The page said “Every write-up of matrixOS this year led with the features.” Too broad: Phoronix (2026-02-11) did lead with the features and did not carry the project’s disclaimer, but It’s FOSS (2026-02-17) carried it in a callout box near the top. The sentence now reads “The February 2026 coverage led with the features, and the largest piece of it did not mention the project’s own disclaimer at all.” A dated note is on the page itself. Found in my own research while writing matrixOS: Fabio Erculiani’s third distro, not reported by a reader — which is why the note names the search rather than a person.

Reporting an error

Write to editor@lxnaydesign.net. Useful mail names the page and the sentence. Mail that also names a source is better, and mail that names a source contradicting me is the most valuable mail this site gets — but a plain “this doesn’t sound right” is welcome too. Chasing down the source is my job.

If you would rather not be credited, say so and you will not be. If you would like to be credited under a handle rather than a name, that works too.