Version history: what changed and when

Fixed

• Browse sorted names by whatever alphabet the reader's browser uses, while the page it arrived as was sorted in English. Slovak and Czech treat ch as a single letter filed after h, so on those browsers Chibados belonged after Cisgender rather than before it, and every tile from that point on sat one place out. The browser noticed the page it was handed did not match the page it would have drawn, discarded it, and drew the whole grid again. All four of the browse orderings are now fixed to English, which is the alphabet the index itself is written in.
• The placeholder that stands in for a plate printed the first half of a letter rather than the first letter. For most terms the two are the same thing. For gala, whose Sumerian is written 𒍑𒆪, the first half is not a letter at all, and the placeholder showed an empty box in its place.
• Neither fault changed a word of any entry. What changed is that the two pages affected, browse and the homepage, now arrive already correct rather than being corrected a moment after they load.

Content

• Plates for the twelve entries that open the catalogue alphabetically, ambigender through butch. Each draws the thing the word is about rather than a person who is it: a balance at rest with equal weights for ambigender, two keys of different cut on one ring for bigender, a jar standing clear of a line of matching jars for aporagender. Drawing the term instead of a body is what keeps an entry from suggesting that a word has a face.
• Four of the twelve are terms belonging to one culture, and their plates hold the objects the sources describe rather than a generic substitute: a two spouted oil lamp for androgynos, a grinding slab and a length of bark for apwint, a ram's horn beside a folded cloth for aylonit, a clay jar and a bound bundle of herbs for babaylan.
• Every plate carries its own description for readers using a screen reader, saying what is drawn rather than what the term means.

Added

• An emblem for the A to Z index: an index roll unfurled and weighted flat, ruled into blank columns, with a straight edge laid across it. The page had been holding a slot for one since it was built. It is an arrangement of what the page is made of rather than a picture of any term filed on it, because an emblem that drew one entry would quietly promote that entry over the hundred and twenty one beside it.

Content

• Every set is printed with its full declension across the five grammatical roles, not the two forms its name is made of. Where a form has two accepted spellings the entry gives both and reports the argument: themself against themselves, hir against zir, and the four cells of xe that no source claims priority over.
• The dates are the point. Thon was proposed in 1858 and sat in a dictionary from 1934 to 1961, which is the answer to the idea that a coined pronoun is a recent thing. Ey was coined by Christine Elverson for a 1975 competition, named after the mathematician who capitalised it in 1983, and put into software on a virtual world in 1991.
• Where the record disagrees with itself, both readings are printed with whoever holds them. Ze is dated to 1864 by one dictionary, to the 1970s and a German source by another, and to a possible Dutch one by a third; none of the three mentions the others.
• Mx is not a pronoun and is here because a title occupies a grammatical slot in front of a name in the same way. It has the strongest paper trail of the batch: first recorded in 1977, a city council accepting it on forms in 2013, the Commons in 2015, two dictionaries in 2015 and 2016.
• One candidate was researched twice and left unwritten. The material for ve/ver came back describing a Bantu language both times, and nothing on file gives the set a declension, a proposer or a date.

Added

• A twelfth group, Pronouns & Forms of Address. The other eleven each name something the sources claim a word means, and a pronoun claims nothing of the kind, so it fitted none of them.

Fixed

• The three commonest pronoun sets were filed on the step of the evidence scale that means a share of the general population, with no figure behind any of them, because that scale measures what people call themselves and a pronoun is not that. All nine sets now sit on the step for things the scale does not measure.

Changed

• Browse and Categories are headed by what they are for rather than by how many things are on them. The counts have not gone anywhere: they sit in the label above the heading, which is still the first line on the page.
• Every heading on the site now says the same thing as the title in the browser tab. Where the two disagreed, a search engine was free to throw the title away and write its own from the page, which is exactly what it does with roughly six titles in ten.
• Nine descriptions were too short to fill a result and two were long enough to be cut off mid-sentence. All of them now sit in the band a result actually shows.
• Six pages, including this one, had no share card at all: a link to them posted anywhere rendered as a blank rectangle. They now carry the site's own card. Every entry, category, collection and kind keeps the card drawn for it.

Fixed

• Two collection titles ran past the width a search result displays and were being cut off before the end of the phrase.

Added

• A row between the page and the footer, on every page, naming KINKSpec and PHOBIASpec. Each card carries that catalogue's own mark and its own colour rather than this site's, because a card in clay would read as a section of this site instead of a different reference.
• Only catalogues that answer are listed. Three more domains in the family are registered and not built, and a link to a site that serves nothing is worse than a shorter row.
• Plates for four of the five kind hubs: concepts, culture-specific categories, expression terms and identities. Pronouns has no entries yet and no page, so it has no plate either.

Fixed

• Pressing the search key on the homepage opened two palettes stacked on each other, one taking what was typed and the other sitting underneath showing the categories. The homepage carries two search buttons, in the masthead and in the opening block, and each was listening for the key on its own account. Only one palette opens now, from either button and from the keyboard.

Content

• Seventy-two entries were rewritten from summaries into articles. Each one separates the history from the counts, says who coined the word and when the record first carries it, and reports the disagreements where a community argues about its own word rather than settling them.
• Twenty-seven entries are new, and they fill the two thinnest parts of the catalogue. Presentation terms went from five to eleven: transvestite, cross-dresser, drag queen, drag king, tomboy and femboy, each of which says plainly that it names a presentation and not an identity. Culture-specific categories went from twenty-seven to forty-one, among them apwint in Myanmar, binabinaaine in Kiribati, khanith in Oman, sipiniq among the Inuit, palopa in Papua New Guinea, nádleehí among the Diné, babaylan in the Philippines, and the galli and the gala of the ancient Mediterranean and Sumer.
• One candidate was researched and then declined. The record for koekchuch, an extinct Itelmen category, comes to about two hundred words and traces entirely to one eighteenth-century Russian colonial report. A page built out of that would misrepresent a living people rather than describe them, so there is no page.
• Roughly thirty claims that the research behind an entry does not support were removed, among them a section on one entry describing a Pacific acronym that appears nowhere in its material, a film dated by no source, and a translation of a word that no page in the folder defines.

Fixed

• Twenty entries printed a Gender Census share beside the article while the article said nobody had counted the word. The survey's own report pages had never been read, so the figure had no page behind it. The reports are now part of the material, and each of those entries either states the count or stops claiming there is none.
• An entry could be seventeen hundred words long, in six sections, and still be filed as a first draft, because the depth tier required a table. The catalogue's own rule is that a table goes in where the material has two axes and nowhere else, so the requirement asked a writer to invent one. Seven entries moved to the top tier without a word changing.
• The check that verifies every year in an entry against its sources could not see a year written in italics, because of how the check read the text. It reported a properly sourced date as unsourced.

Added

• A missing page prints the address that failed and then the entries whose names come nearest to it. Typing nonbinary finds non-binary, faafafine finds fa'afafine, genderflud finds genderfluid and genderflux, and an address with no near match says so rather than offering the least bad guess.
• The match is made on letter pairs rather than on exact spelling, so it is close to indifferent to the two things that break lookup in this subject: romanisation and hyphens.

Added

• A section on every entry setting out the six steps: what the word reached, what the step required, and which datasets define it. Every step is expandable, including the ones the word did not reach, so a reader can see what it would take to move up one.
• The step rides in the facet row beside the origin and the category, in the colour of the step. Origin is a kind and gets six distinguishable hues; the step is a quantity, so it runs strongest to faintest and the fade is the argument.
• A provenance block under the opening paragraph of every entry: the dated first use at the size a headline figure deserves, the coiner where one is on record, and the count where anybody has made one. It replaces five grey rows in the sidebar, where a reader met the one thing this catalogue prints that the other glossaries do not only after finishing the article.
• The method page explains the whole scale, with the same ladder the entries carry.

Changed

• Chips, counts and mono labels were set in faint ink at the smallest size on the page with the widest letter spacing, which is three legibility penalties at once. They now read at the weight of the facts they carry, and the two faintest ink values were lifted in both themes.
• Subheadings inside an entry were a tenth of an em smaller than the section headings above them, at the same weight and colour, which is a difference in rank nobody can see. A subheading is now a fifth smaller, indented, and led by a short rule.
• The method page opens beside the site's own plate rather than as a wall of argument, and the five kinds of entry are a grid of cards instead of five faint chips with a paragraph hung under each.

Fixed

• The entry for cisgender named no coiner. The field held an etymology instead, while the body of the same entry carried a section on who actually coined the word.
• The captions under the plates on category and collection pages labelled a picture that was plainly an emblem of the category, and next to a placeholder they labelled nothing at all.

Added

• Four kind hubs, one for each kind with entries filed under it. Each carries a two paragraph argument, four questions of its own, its own share card and its own plate, written in a file beside the categories rather than hard coded in the front end.
• The kinds are in the sitemap and in the file the answer engines read. A hub with no entries filed under it is left out of both until the first entry lands.
• The catalogue browser gained a filter for the kind of term, and the classification rail on an entry now links: the kind goes to its hub, the category to its own page, and the origin and the step to the browser filtered by them.

Changed

• The footer stopped ending two of its columns with an arrowed link. An arrow in a colophon is a button pretending to be typography; both are now ordinary rows of the list they belong to, separated by a hairline.

Changed

• Transgender, non-binary, cisgender, trans woman, trans man, intersex, two-spirit, hijra, eunuch and butch are rewritten at full length, with the history and the counts separated from each other and the neighbouring terms drawn in.
• Every entry now records how far its own body has been taken, counted from the text rather than asserted, so the queue of entries still to deepen is a field rather than a memory.

Fixed

• Four claims in the existing bodies were not supported by the material the entries were written from and are gone: an act of parliament named on hijra, a punctuation convention and a dictionary dating on transgender, and a claim about how many butch people are cisgender that nobody has counted.
• Two clinical outcome figures inherited on trans woman are removed. The catalogue does not carry outcomes.
• The entry for intersex dated the English word to a dictionary that carries no such date. It now credits the geneticist who coined the longer form in 1917, who does.

Added

• Plates for five category hubs: agender and neutral genders, the demigenders, the distinct genders, gender expression and the fluid genders. Each is a still life standing for the group, because drawing one of a category's terms on its hub quietly promotes that term over the others filed beside it.
• Every plate carries its own description for readers using a screen reader, saying what is drawn rather than what the category means.

Fixed

• One plate came back with a colour code lettered across the drawing, another with the same instrument drawn in two colours. Both were the instructions leaking into the picture rather than governing it, and both are now impossible: the colour is named in words, and the drawing carries one colour and no more.
• The generator stamps a small mark into the corner of everything it returns. It is removed from every plate before the plate reaches the site.

Added

• A graphite portrait plate on the homepage, beside the paragraph explaining what the catalogue is for, and on the share card. Every entry, category and collection already carries a written prompt for its own plate; this is the first one drawn.

Fixed

• The share card for an entry printed a grey dot and no label where it should have said where the word came from. It was keyed on a scale this site stopped using, so every entry looked identical to a reader scrolling a feed.
• The footer column headed The Record was four blank rows counting zero, for the same reason. It now lists the five origins the catalogue actually holds, with their counts.
• The homepage said gender terms come from four places. There are six, and five of them are in use here.

Content

• Eighty-two entries, one for every term in the reference list the catalogue is built from: the binary and the words for crossing it, twenty-three gender categories belonging to particular cultures, the umbrella terms, the fluid and the partial and the multiple, and the microlabel tail.
• Depth follows the record rather than a house length. A term with population figures, a dictionary attestation and a professional-body definition runs past a thousand words; a term whose entire documentation is two paragraphs on a community wiki runs to three hundred, and the entry says so. Padding the second to match the first would have meant inventing the difference.
• Every entry prints where its word came from: the year, the person and the publication where those exist, and a plain sentence saying nobody has dated it where they do not. That last case is the commonest one in this subject and no other glossary prints it.
• Eleven categories, read out of the sources rather than drafted in advance. The ten planned earlier failed against the catalogue that actually exists: two of them had no entries at all and one held a third of everything.
• Eight collections, each making an argument a filter cannot: the pairs readers most often mix up, the figures that get misquoted and why, the words whose meaning has moved, the categories that resist translation, and the Pacific grouping named by the communities in it rather than by this site.

Changed

• The scale the site sorted by is gone. It ranked terms by how much literature stood behind them, which does not transfer to this subject: a gender term is not more real for having been studied. In its place the catalogue sorts by where a word came from, which is a kind rather than a rank, and nothing in the interface now implies one origin is better than another.
• Numbers are labelled by what they measure. One tier means a share of a general population and the rest mean a share of the people who chose to answer a community survey. Reporting the second as the first is the commonest error made about this subject, and the site now refuses to make it in its own tooltips.

Fixed

• Three dates in one entry had been written without a source behind them. A sweep of the whole catalogue found unsourced years in seven more. All are cut or traced back to a source, and a check now runs on every build so that a year cannot enter an entry unless it is in the material that entry was written from.
• Search now matches the way names are actually spelled. Diacritics, apostrophes and hyphens are collapsed on both sides, so faafafine finds Fa'afafine and nonbinary finds non-binary.
• The page shown when a term is not found offered an example from a different subject entirely.

Added

• Every route a reference catalogue needs: an entry page, category and collection hubs, a browse view with facets, an A to Z index that works with JavaScript switched off, and a search that reads a build-time index.
• The design system. Clay is the accent, at #B5623C in the artwork and #AD5A35 in the interface, and it was chosen because it sits outside gender colour coding entirely: not pink, not blue, not rainbow. A reference work that signals a position in its own chrome has argued the case before the reader reaches a sentence.
• The mark is a plotted term: an axis with no end stops and one dot standing above it, off the midpoint. A bounded line with a point on it draws a spectrum between two poles, and that is the most common wrong picture of the subject.

Content

• No entries yet. What the site is showing is a placeholder corpus, and every page that describes the catalogue will be rewritten against the real one. Pronouns are written first: twenty sets, each with a full declension table and a dated first attestation, which is something no existing resource provides consistently.