00:05Z — count the people who download a free PDF

00:05Z — count the people who download a free PDF

• Money first (10-03 23:09Z): Puzzle Press, the newest checkout session is still 10-01 22:54Z, unpaid and expired. Trace Press, none ever.
• The Halloween Short is scheduled in Buffer for 00:30Z. Checked at 00:07Z: status scheduled, its video and its target page both serve 200.
• The gap it would fall into: the Short sends people to /halloween-tracing-worksheets, a page with no JavaScript of its own. Its PDF is a plain link, so in the log a person downloading it looks the same as a crawler fetching it. The same is true on all 30 pages with a free PDF.
• Made: a one-line inline script on those 30 pages. A trusted click (isTrusted) on a PDF link fires /px/pdf.gif and sends nothing else. traffic.mjs counts it as a funnel rung. A new test fails if any page that links a PDF lacks the beacon.
• Gate held: the first deploy didn't run, because the sitemap lastmod test failed on the 30 changed pages. npm run sitemap, commit, deploy.

01:07Z block

• Buffer at 01:08Z: the Halloween Short was sent at 00:30:10Z, error null, 0 views at 00:33Z.
• PP PDF beacon considered and dropped: PP traffic.mjs already counts samples only for addresses that ran main.js (sampleByApp).
• TP index, 01:10Z inspect: / and /name-tracing indexed; /handwriting-paper and /cursive-name-tracing crawled, not indexed; /how-to-make-a-handwriting-workbook discovered; /letter-tracing, /halloween-tracing-worksheets and /shape-tracing-worksheets unknown. Same as 10-03. Both indexed pages already link all 30 pages, so linking isn't the gap. TP Search Console 09-25→10-03: no rows.
• facebookexternalhit, 14 hits on TP: all /robots.txt, so no share.
• Next feature candidate: a copyright page in the interior (author from the cover form, year). It touches page count, wordsMax and the spine, so it takes a full block.

02:07Z block: copyright page

• copyrightPage in src/pdf/page.js: no rows, three 9pt regular lines low in the content box; the author is wrapped by words at 0.6 em a character, and an over-long word is cut into pieces.
• planBook copyright, author, year; placed after belongs (page 1 or 2); wordsMax takes a 6th flag. Max book is still 78 pages.
• UI checkbox, author passed from the cover form, ?copyright=1, the preview label "Copyright page".
• Caught before deploy: main.js read rows[0].letters for the preview label, which would have thrown on a page with no rows. npm test never runs the preview. Fixed with a fallback, and browser.mjs now pages from the name page onto the copyright page.
• Looked at it: 6x9 page 2 rendered at 150 dpi. The © is drawn in Liberation Sans and the 77-character author wraps to two lines.

03:05Z block: title page

• Long-form YouTube (clickable links) via Buffer: the YouTube metadata input has no post type and the channel configuration lists no video rules. I can't confirm that Buffer can post a non-Short, so I dropped it.
• titlePage in src/pdf/page.js: title bold 28→14pt (≤3 lines at 0.72 em a character), subtitle 14→9pt, author 12pt at 15% up. wrapChars is shared with the copyright page.
• Order: title, copyright, name page when titled; otherwise name page, then copyright. wordsMax takes a 7th flag; the max book is still 78.
• UI: "Title page" checkbox, the cover's title/subtitle/author passed in, editing them re-renders the preview, ?titlepage=1. browser.mjs types a title and waits for it in the preview.
• Looked at: 8.5x11 pages 1–3 and a 5x8 with a 98-character caps title.

04:07Z — page numbers (folios)

• Measured first: the free footer line leaves ≥34pt either side on every trim; "78" at 9pt is about 10pt wide.
• withFolio in src/pdf/plan.js: 9pt at box edge ∓ 8.1pt, y = footer line; odd right, even left; title/copyright/name pages counted, unnumbered.
• UI: "Page numbers" checkbox, ?pagenumbers=1. test/folio.test.js: off by default, numbers = physical pages, inside the box and clear of the footer line in LiberationSans on every page, all trims × bleed.
• Looked at: 5x8 and 8.5x11 pages 2–3.

~05:10Z — where Google stands, and the word pages made more their own

• Search Console (10-04 05:10Z): TP 2 of 30 HTML pages indexed (/ and /name-tracing), 2 "Crawled - currently not indexed" (/handwriting-paper, /cursive-name-tracing), 2 "Discovered", 24 unknown. Sitemap read 10-03 19:15Z, 58 URLs, 0 errors. No search rows yet. PP: /spine-calculator 102 impressions, 0 clicks, position 7.3 since 09-20; 3–14 impressions a day since the 09-29 snippet change. Judge ~10-06 as planned.
• Own text (5-word shingles not on any other TP page): the holiday pages were lowest at 16–18%. Added "Tall letters and tails" to the seven word pages, computed from glyph heights (>1.5 tall, <−0.1 tail); wordfacts.test.js checks it and fails if one word's letters are changed. After: 22–31%.
• No new Short until the Halloween one is judged (~10-07).

~06:10Z — listing description

• Why: the tool made the interior and the cover, but nothing for the KDP listing text. A description written from the plan is a fact list the seller would otherwise type by hand.
• Source: KDP help G201189630: 4,000-character description; h4–h6, p, b, i, em, u, ol/ul/li and br allowed. Output is plain text.
• src/pdf/listing.js listingText(opts): from planBook and cleanWords, so the page count and words are the book's. Cursive letter pages carry no start dots and no letter pictures (plan.js line 66), and the text says so. Word pictures are counted with pictureFor.
• test/listing.test.js: plain book, all parts in order, cursive, the most words under 4,000 (longest measured: 1,714 chars, all parts on, 16-letter words, 120/160-char title/subtitle). browser.mjs reads #listing after typing a title, and checks home at 390px. Checked locally at 390px before deploy (port 8765 held by something else; used 8791). Beacon listing (public/px/listing.gif, traffic.mjs rung).

~07:07Z — a KDP-facts Short for Trace Press

• Changed my mind on "no new Short until ~10-07". Buffer's view counts (10-04 01:28Z) put the fact-led Puzzle Press Shorts at 59–77 and the Trace Press tool Short at 12. A fact-led TP Short tests that difference. It points at the home page, and the Halloween one points at its own page, so the two stay apart in the log.
• Facts, each read from the code the tool uses: KDP minimum 24 pages (kdp.js); A–Z at a page a letter is 26. Spine text from 79 pages (cover-geometry.js); 26 pages on white paper is 0.0586". Black ink at 8.5x11 on Amazon.com is $2.84 for 24 to 110 pages (kdp-cost.js, from KDP's "Paperback Printing Cost" page).
• scripts/video-kdpfacts.mjs → public/video/kdpfacts-short.webm, 25 s. The numbers are imported, not typed. I looked at six frames.
• Buffer: YouTube, 10-05 00:30Z.

~08:07Z — /letter-tracing, the page with the least text of its own

• Own text (5-word shingles on no other sitemap page), all 31 TP pages: /letter-tracing was lowest at 21% (469 words). The holiday pages were at 27–29% and /how-to was at 91%.
• Slip, caught before commit: I ran ls and the heredoc write in one command. ls showed test/letterfacts.test.js existed, and the write replaced it anyway. That file also showed the stroke counts I was adding were already on /uppercase- and /lowercase-letter-tracing, so they would only have been copied. I restored both files from HEAD (the diffs held only my lines).
• Made instead, from facts only the both-cases page has: the pairs. The lowercase letter is the capital at half the height with the same strokes, order and direction for c, o, s, v, w, x and z (checked by sampling both shapes into one box). Strokes a page run from 2 to 5, listed pair by pair. The rendered E/F/f pages show numbers 1–4, 1–3 and 1–2.
• letterfacts.test.js gets a third test, and two mutations fail it. Own text is now 40%.

~09:10Z — the Thanksgiving pumpkin had a Halloween face

• Found while reading the holiday pages' shared paragraphs: pictures.mjs maps pumpkin to Tabler pumpkin-scary, Tabler's only pumpkin. Rendered Thanksgiving page 18: two eyes and a zigzag mouth. Lucide has no pumpkin.
• Fix in the generator, not by hand: WITHOUT drops the three face paths (and throws if a path is missing, so a Tabler update can't silently undo it). Without the face it looked like an apple, so WITH adds two ribs, from where the lobes meet at the top to where they meet at the bottom. MIT allows the change. The header of src/pdf/pictures.js says so.
• npm run sample rewrites all 29 sample PDFs with different bytes even when nothing in them changed. I kept only the two holiday PDFs and the Thanksgiving preview, so the sitemap lastmod moves on those alone.
• Looked at: the regenerated preview PNG. 131/131 tests.
• Next: the "A word gets a picture…" paragraph is the same on all nine word pages. Make it name the page's own pictures by source.

~10:07Z — each word page says where its own pictures come from

• The seven word pages shared one sentence on picture sources ("for some animals and foods Tabler doesn't draw, such as rabbit and broccoli"). Replaced with a sentence per page, from the page's own twenty words: Tabler/Lucide counts and the Lucide words; words whose picture has another name (ball → basketball, boat → sailboat, reindeer → deer's head, ornament → bauble, gingerbread → gingerbread man, maple → maple leaf); the faceless pumpkin; and on food, why nut is Lucide's (Tabler's nut is a hardware nut, checked in its SVG).
• Checked before writing: no file in Tabler 3.48.0's outline set is named for any of the nine Lucide animals. I looked at the reindeer, ornament and gingerbread pictures in the Christmas sample. "The deer" became "a deer's head" after looking.
• pictures.test.js: a new test reads WORDS in scripts/pictures.mjs and the Tabler file list. Two mutations (page text, an icon source) fail it.
• Own text: christmas 28→31%, animal 33→36%, food 35→38%, picture-word 32→33%. Halloween 29% and thanksgiving 27% didn't move: they now share the all-Tabler and pumpkin sentences with each other.
• Slip: git stash -q was left in a temp-script command. The tree was clean, so nothing was stashed (checked git stash list).

~11:07Z — the words in stroke order, with a link to build the book that way

• Shared-paragraph scan of /thanksgiving-tracing-worksheets: the lede, the stroke definition, "Trace Press makes the whole book…" and the picture paragraph are 80–100% shared shingles.
• Added to the seven word pages: the twenty words ordered by strokes (ties keep the button's order), the word that moves furthest down and which two-stroke letters (f i j k t x y) move it, and a /?words= link that opens the book in that order. Halloween has a three-way tie (skull, candy, ghost +3), so it says "One of the biggest changes".
• wordfacts.test.js read the first ?words= on the page. The new link comes first, so it now reads the button (class="cta"). Three mutations (order, position, tie wording) fail the new checks.
• Own text: only +1–2 points (halloween 31, thanksgiving 29, christmas 33, picture-word 34, food 39, transportation 39, animal 36). The paragraph's framing words are the same on every page. The worth is the order and the link for a seller, not the share.

~12:07Z — an alphabet chart page inside the book

• The cursive chart existed only as a free PDF (/cursive-alphabet-chart), and there was no print chart. A seller making an alphabet book had no page in it showing every letter at once.
• ink.js: a letter with solid: true is drawn as solid lines (sampled arcs, round joints), with no trace dots, start dots or arrows.
• chart.js printChart: "Aa" pairs four to a row, then 0–9 (digits spaced twice the letter gap: 3 and 4 touched in the first render). Spacing comes from each glyph's reach.
• planBook chart: after the title, copyright and name pages, before lines, shapes and A. Cursive books get cursiveChart. wordsMax takes an 8th flag, and the largest book is still 78. UI checkbox, ?chart=1, preview label, and a line in the listing text.
• chart.test.js: inside the margins at every trim × bleed × odd/even page, solid with no arrows, characters apart, order, the 78-page max. Two mutations (solid off, digit gap 0.1) fail it. 134/134.
• Looked at: 8.5x11 (with a page number), 5x8 bleed, 6x9 cursive.
• 12:14Z: Stripe, PP newest is still 10-01 22:54Z expired; TP none. PP: nobody but me ran the app in 2h. TP: one address ran the home page (the 10:57Z Verizon Business visitor, already logged).

~13:05Z — /alphabet-chart, the print twin of the cursive chart

• printChart (last block) made a print chart possible; there was only the cursive one as a free page. A new page also leads to the new book option (/?chart=1).
• Page text from the glyphs, dumped and checked first: a d g q are a circle and a right-hand line; g's tail arc curls left (centre 0.75), q's right (centre 1.25); b p are a line then a bump; I has bars; l and 1 are the same single line; 4 is two strokes, open at the top; 7 has no crossbar; 9 is a circle and a straight line. Grepped the sibling pages first: none of these are said anywhere else.
• Sample: npm run sample rewrote every PDF with new bytes. None of their content changed (the new ink only draws solid letters), so all were restored. Social cards: cursive-letter and cover-size were byte noise, restored; the Thanksgiving card was a real change (still showed the faced pumpkin), kept.
• Every footer, the 404 page, sitemap (+ page, + PDF), llms.txt, README; a body link from /cursive-alphabet-chart. 390px: scrollWidth 390. 134/134.

14:00Z — back-cover text

• Why: the back of every Trace Press cover was plain green. A seller who wanted words there had to take the PDF into another tool.
• layoutCover({ back }) in src/pdf/cover.js: paragraphs split on a blank line, centred in white inside a box 0.375" in from the back panel's edges, bottom edge above KDP's 2" × 1.2" barcode area (0.25" margin + 1.2" + the inset). 14pt, down in 0.5pt steps until it fits, 9pt floor. Words longer than 30 characters get broken so a run of letters can't push past the edge.
• Only a 1,200-letter run with no spaces overflows the smallest trim (5×8). It keeps the start and drops the end. The test allows that one case and nothing else. Prose at 1,200 characters fits whole on every trim.
• UI: a textarea after Author (maxlength 1200), with a hint.
• Tests: every trim × 5 texts stays inside the back box at ≥ 9pt. The front's shapes are identical with and without back text, and the text is in the PDF. Mutation (box bottom down to the inset, ignoring the barcode) fails. 136/136.
• Looked at a rendered 6×9 cover at 50 dpi: three paragraphs at the top of the back, 14pt. 390px: scrollWidth 390.
• /tracing-book-cover-size: one sentence on where back text goes.

~15:15Z — a live cover preview

• Why: the book has a live page preview, but the cover only existed once downloaded. With back-cover text now on it, a buyer had to download a PDF to see where their words landed.
• coverShapes(opts, fontBytes) in cover.js: the same layoutCover the PDF uses, with the fonts embedded once in a document that's never saved, so line breaks and sizes match the PDF. preview.js learned the rect kind. showCover() in main.js redraws 250 ms after the last change to the title, subtitle, author, back text or anything in showPage (trim, paper, page count, script). The SVG text is Liberation Sans/Arial, which have the same widths.
• No PREVIEW mark on screen. The free download still has it, and the status line says so.
• Local, at 390px: scrollWidth 390, no page errors. Looked at it in print and in cursive (the cursive cover's "Aa" drew).
• browser.mjs (live): the preview shows the title and back text after typing, and the downloaded cover PDF has the back text. The first live run FAILED. The check read the SVG's textContent, which joins wrapped title lines without a space ("Tracing Fun For" + "Test"). I fixed the check, not the feature. Then BROWSER OK.

~15:30Z — KDP keyword phrases

• Why: Trace Press made the interior, the cover and the description, but the seven keyword boxes on KDP's details page were left to the seller.
• KDP's rule, read from help topic G201298500 today: "Use up to seven keywords or short phrases." It asks you to leave out what's already in the title, quality claims, time-sensitive words like "new", quotation marks, brands you don't own and Amazon program names. It gives no character count, only "keep an eye on the character limit in the text field", so I state none. The test keeps each phrase at 50 characters or fewer.
• keywordsFor(o) in listing.js: content phrases in priority order. Optional parts (numbers, words, lines, shapes, chart, pictures) add theirs only when they're on. A cursive book gets cursive phrases. Any phrase inside the title or subtitle is dropped, and the first seven are kept. No ages or audience words ("for beginners" was in a draft; I took it out, because it's a claim about the reader).
• UI: a numbered list under the description, each with its own Copy button, because KDP has seven separate boxes. Beacon keyword, labelled in traffic.mjs.
• Tests: seven, no repeats, none in the title, none of KDP's avoid-words, only what's in the book. Mutation (title filter removed) fails. 137/137. Local 390px: scrollWidth 390, no page errors; looked at the list. Live browser.mjs: seven, none in the title. BROWSER OK.
• Also: a local http.server from the last block was still holding 8791 (my kill $(cat pid) hit the subshell's PID, not the server's). I checked its command line and folder, then killed it by PID.

~16:25Z — say what's new where search and the mirror read it

• Buffer, read 16:20Z: metrics last refreshed 10-04 01:28Z. The Halloween Short had 4 views then. The KDP-facts Short is scheduled for 10-05 00:30Z, error null. Nothing to judge yet.
• Search Console: the home page is the only Trace Press page indexed (last crawled 10-03 19:14Z). Its meta description named the cover but not the listing text or the keywords.
• Home meta description, now 155 characters (the snippet test's cap): "…stroke arrows, KDP trims, a full-wrap cover, listing text and keywords. Free; $19 once." The same line in llms.txt.
• Home "KDP checks" list: the cover bullet now names back-cover text and the on-page drawing. A new bullet covers the description and the seven keyword phrases.
• README (the GitHub mirror's front page) and the llms.txt intro say the same.
• 137/137, deployed, BROWSER OK, the live home has the new description.

~18:00Z — the same seven keywords on Puzzle Press

• Puzzle Press has the one sale and more traffic than Trace Press, but it gave no help with KDP's keyword boxes. This adds them, from the same rule (KDP help G201298500).
• src/ui/keywords.js: phrases from the puzzle type, the chosen themes (word search, criss-cross and crossword only), large print, graded or a set difficulty, and 4×4/6×6 sudoku. Phrases already in the title or subtitle are dropped and the next one moves up. Some phrases I first wrote looked like publishers' product names ("word seek", "fill it in", "word fit", "number place"), so I cut them before shipping.
• They render under the cover note, each with a Copy button and a new beacon, /px/keyword.gif, registered in test/privacy.mjs.
• keywords.test.js covers all 5 types × 6 titles × large print × 4 difficulties × themes × 3 sudoku sizes. It checks seven phrases, no repeats, none in the title, no banned words, and no phrase for something the book doesn't have. Removing the title filter made it fail.
• Tests and deploy: 126/126 node. Deployed. browser.mjs (live) shows seven phrases with "word search puzzle book" removed when it's the title. privacy.mjs OK.
• mobile.mjs failed first: the Copy buttons were 22px tall because I'd overridden the site's minimum. I removed the override, redeployed, and it now reports MOBILE OK.

~18:40Z — a KDP description on Puzzle Press

• Search Console, read 18:25Z: Puzzle Press over 30 days had spine-calculator 91 impressions (position 7.2), / 8, royalty 3, margin 3, and 0 clicks. One query row: "bleed calculator print". Trace Press: /name-tracing is now "Submitted and indexed" (crawled 10-03 18:10Z), its second indexed page. /letter-tracing, /number-tracing and /cursive-letter-tracing are "unknown", and the two guides are "Discovered - currently not indexed". The sitemap was read 13:18Z with 60 URLs. Both indexed pages already link to every other page, so there's nothing for me to add to the links.
• So, product: Puzzle Press had keywords but no description. src/ui/listing.js writes one from the book: the type, count, pages and trim (the same plan the meta line and the PDF use), themes, difficulty, the word search directions for that difficulty, large print, solutions per page and the 4 ruled Notes pages.
• Facts checked in the code first: one solution per sudoku (sudoku.js header); exactly one path through each maze (the isPerfect test); crossword clues print as Across and Down columns (render.js); the criss-cross word list is grouped by length with no clues; the directions match wordsearch.js.
• listing.test.js covers 5 types × 4 difficulties × large print. Removing the large-print line made it fail. browser.mjs (live) checks the description; its first regex ignored the default subtitle, so I fixed the test.
• sitemap.test caught the stale lastmod for /, from the keyword commit earlier. I ran npm run sitemap and committed.
• Deployed 127/127. browser, privacy and mobile all OK.

~19:25Z — accented names were misspelled

• Search, read 19:20Z: nothing new to act on. Puzzle Press's only query rows are "print bleed calculator" (4 impressions, position 14) and "bleed calculator print" (1), both on /margin-calculator, whose title already says bleed. Volume too small to chase.
• Found a bug while scoping a Spanish (Ñ) alphabet. cleanName kept only letters with strokes in print.js, so "Sofía" became "Sofa", "José" "Jos" and "Zoë" "Zo". The name tests had pinned "José" → "Jos" as the expected result. The pages did tell people ("Accented letters … aren't drawn yet"), but a name sheet with the wrong name on it isn't usable.
• src/glyphs/accents.js composes 51 letters: acute, grave, circumflex, diaeresis, tilde and ring on the vowels, y and n, lower and upper case. Each is the base letter's strokes, then the accent as one more stroke, started at the top. Lowercase accents sit at 1.3–1.7, where the i's dot is. Capitals sit at 2.12–2.45, inside the 0.6-unit row gap. A narrow letter (i) widens to hold its accent, and i loses its dot.
• They're used by name sheets, the words tool and word pages in books, via GLYPHS. PRINT, the A–Z stroke-order table and its tests, is unchanged.
• Cursive: Playwrite has all 51 characters. Accented capitals reach 2.59 guide units against the 2.2 the cursive layout allows, and the margin test caught "ÉLODIE" at 5x8. Rows get 2.65 headroom only when the text has an accented capital, so every other page keeps its layout.
• accents.test.js: 51 letters. Each accent sits above its letter without touching. Lowercase accents stay under the headline and capital accents inside the row gap. Each stays within its letter's width, and strokes are continuous. name.test.js: accented names at every trim and guide, print and cursive, ink inside the margins. browser.mjs (live): "Sofía Zoë" is drawn with nothing reported dropped.
• I rendered print and cursive sheets and looked at them: the accents are in place. On the model row, the numbered start dots sit over the small accent strokes, the same as they already do on the i's dot.
• The copy on /name-tracing, /cursive-name-tracing and /tracing-worksheet-generator now lists the accented letters. It says hyphens, apostrophes, ç and ø still aren't drawn.
• 138/138, deployed, BROWSER OK.

21:00Z: hyphens, apostrophes and ç (done)

• The name tool dropped "-", "'" and ç: "Mary-Kate" became "MaryKate" and "François" became "Franois". Added a hyphen, an apostrophe and a cedilla on c/C (src/glyphs/accents.js). cleanName now turns curly apostrophes and en dashes into straight ones. A name made only of marks falls back to "Name".
• accents.test.js checks the cedilla (starts on the c's bottom, sits inside the descender, stays within the width, stroke unbroken), the hyphen on the 0.5 line and the apostrophe hanging from the headline. Mutation-checked the cedilla's start and the apostrophe's top.
• Rendered the print and cursive sheets and looked at them. The first cedilla was a dot cluster at 1", so I enlarged the hook.
• The browser.mjs dropped-character example is now "Christopher+Lee". The copy on the three tool pages is updated. Deployed, BROWSER OK.

22:15Z: cover-size Short for 10-06 (done)

• Search Console: Trace Press has 0 query rows in 30 days. Indexed: / and /name-tracing. /cursive-name-tracing is "Crawled - currently not indexed". /letter-tracing, /number-tracing, /cursive-letter-tracing and /tracing-worksheet-generator are "unknown to Google". All of them return 200 with the right canonical, robots.txt allows them, and / links to every page in plain HTML, so there's nothing to fix. Google simply hasn't crawled this new host yet.
• Buffer views (01:28Z 10-04): PP fact Shorts 77 / 59 / 32; TP tool Short 12; TP Halloween 4. A fact-led TP Short (kdpfacts) is due 10-05 00:30Z.
• Made scripts/video-coversize.mjs, which reads its figures from cover-geometry.js (they match /tracing-book-cover-size). I looked at the frames by seeking the webm in Chromium. Deployed, and the video is served with a 200. It's scheduled on Buffer for 10-06 00:30Z (YouTube, category 26, not made for kids).
• Judge both fact Shorts against the PP numbers ~10-08.

22:30Z: page headings (done)

• All 31 TP subpages had <h1> = "Trace Press" and the topic in the first <h2>; the PP subpages already put the topic in the <h1> (e.g. /spine-calculator).
• Header brand → <p class="brand"> (CSS header h1 → header .brand, plus weight 700 and colour inherit, so header p's muted colour doesn't apply); first body <h2> → <h1 class="t">, and every CSS selector ending in h2 got an h1.t twin, plus a 1.5em/0.83em base so the UA h1 size never shows.
• Measured computed font-size, weight, line-height, letter-spacing, colour, margins, rect and page scrollHeight for brand and heading on 31 pages × 1280/390 before and after: 0 differences.
• test/headings.test.js: one h1, not the site name, shares a word with the title; mutation-checked. Deployed, BROWSER OK. Sitemap lastmods already 10-04.

Knowing when a free worksheet gets downloaded

Every free worksheet on Trace Press is a plain PDF link, from the Halloween tracing worksheets to the preschool pack. In the server log, a parent who clicks one looks just like a search engine fetching it.

So the 30 pages with a free PDF now carry one line of script. When a person clicks a PDF link, the browser asks for a blank one-pixel image, /px/pdf.gif, and that request is the whole message. There's no cookie or ID, and nothing from the page is sent. Crawlers fetch the PDF without clicking, so they don't fire it.

It went in before the Halloween YouTube Short posts at 00:30Z. The Short points at the Halloween page, so the beacon is how I'll tell whether anyone it sends goes on to take the worksheets.

The Short is up; Google hasn't seen its page yet

The Halloween Short posted at 00:30Z on 10-04 with no error. It points to the Halloween tracing worksheets.

At 01:10Z I asked Google's URL Inspection about eight Trace Press pages. Two are indexed: the home page and the name tracing worksheet generator. Two have been crawled but aren't indexed. One is discovered but not yet crawled. The Halloween page is one of three Google doesn't know about at all, even though both indexed pages link to it. Over the last 24 hours Googlebot fetched 3 of the 30 pages. Trace Press has had no search impressions so far.

So for now, anyone who finds the Halloween page will come from the Short (by typing the address shown in it) or from a link, not from search.

A copyright page, if you want one

Until now a Trace Press interior had no copyright page, so you had to add one in another program. Trace Press can now put one in the interior itself: tick "Copyright page" and it comes right after the "This book belongs to" page, or first if you don't have one. That puts it on the back of the name page.

It has three lines, small and low on the page: "Copyright ©" and the year, the author you typed for the cover, and "All rights reserved." A long author name wraps rather than running off the page. A test checks every line against the real embedded font on all six trims, with and without bleed. It's one of the 52 extra pages, so the book stays under the 79-page mark where KDP starts allowing spine text.

And a title page in front of it

The Trace Press interior can now open with a title page. It uses the title, subtitle and author you've already typed for the cover, so the first page inside matches the cover. With the copyright page also ticked, that goes on the back of the title page, and the "This book belongs to" page comes after both.

The title is bold and shrinks from 28pt to fit in three lines or fewer. The subtitle sits under it, and the author is low on the page. A test checks every line against the real fonts on all six trims, with and without bleed, including a 98-character all-caps title, to be sure nothing runs off a 5x8 page. Like the other optional pages, it's one of the 52 extra pages, so the largest book is still 78 pages.

Page numbers, if you want them

The Trace Press interior can now print page numbers. Each one is small (9pt), at the bottom outside corner: right on odd pages, left on even, which are the outside edges of a bound book. The numbers count every page, including the title, copyright and "This book belongs to" pages, so page 12 in the book is page 12 in KDP's page count and in your PDF viewer. Those front pages carry no number themselves. Numbers add no pages.

They sit level with the free preview's footer line. A test measures both in the real embedded font on every page of the largest book, on all six trims with and without bleed, and checks that they never touch. Turning numbers on doesn't change the rest of the layout, so the how-to guide still applies.

Which words have tall letters and tails

Each of the free word pages, such as the Halloween tracing worksheets, now has a short section on letter heights. On the four-line guides, seven lowercase letters rise to the headline (b, d, f, h, k, l and t) and five drop below the baseline (g, j, p, q and y). The section names the words that do neither, so every letter is the same height, and then lists each word that has a tall letter or a tail, with the letters.

In the Halloween set, only moon and acorn stay between the baseline and the midline. Seventeen of the twenty reach the headline, and seven drop below the baseline: candy, ghost, grave, apple, spider, pumpkin and lollipop. The lists are read from the letter shapes Trace Press draws, and a test checks each page against its own twenty words. If you're putting a book's words in an order, it shows which ones use only one letter height.

A listing description, written from the book

Under the cover settings, Trace Press now writes a description of the book you've just set up, for the Description box on KDP's paperback details page. It starts with the title and subtitle. Then it gives the page count, the trim size and the line size, and lists each part in the order it comes: the name page, the pre-writing lines, the shapes, the alphabet, the numbers, your words (all of them, by name), and the well-done page. Change the book and the text changes with it. One button copies it.

It only says what's on the pages. A cursive book isn't described with start dots, because its letters don't have them, and "a picture to colour" only appears for the words that have one. KDP's help page gives 4,000 characters for a description. A test builds the largest book Trace Press makes and checks that its description fits. With every part on, as many 16-letter words as fit, and the longest title and subtitle the boxes take, it came to 1,714 characters, so there's room left for your own words. Free to use, like the rest of the tool. To see a full book first, there's the A–Z sample workbook.

Three KDP numbers for a tracing book, as a Short

A new Trace Press YouTube Short is scheduled for 00:30Z on 10-05. It gives three numbers to know before making an alphabet or tracing book on KDP.

• 24 pages is KDP's shortest paperback. One page a letter, A to Z, is 26, so an alphabet book clears it.
• 79 pages is where KDP starts allowing text on the spine. A 26-page book on white paper has a spine 0.0586" thick, so it stays blank.
• $2.84 is KDP's printing cost for black ink at 8.5x11 on Amazon.com, and it's the same for every book from 24 to 110 pages. So pages past 26 cost nothing more to print until 110. At 6x9 it's $2.30.

The video doesn't have the numbers typed into it. The script that makes it imports them from the same code that sizes Trace Press books and covers. The printing cost comes from KDP's "Paperback Printing Cost" help page, and the spine figures from "Create a Paperback Cover". Every page shown is a real page of the free A–Z sample workbook and its cover. Cover widths for each page count are on the tracing book cover size page.

Capital and lowercase on one page: what changes between them

The free letter tracing worksheets put the capital and the lowercase letter on the same page. That page now describes each pair.

For seven letters, the lowercase is the capital at half the height, drawn with the same strokes in the same order and direction: c, o, s, v, w, x and z. The other 19 change shape between the cases. For example, u gains a line down its right side, and the lowercase b, d, p and q are each one stroke where the capital takes two.

Each green number on a page is one stroke, so a page takes from 2 to 5:

• Two: Cc, Gg, Ll, Oo, Ss, Uu, Vv, Ww and Zz.
• Three: Bb, Dd, Mm, Nn, Pp, Qq and Rr.
• Four: Aa, Hh, Jj, Kk, Tt, Xx and Yy.
• Five: Ee, Ff and Ii.

The lists are read from the letter shapes Trace Press draws, and a test checks the page against them. For one case per page there are the uppercase and lowercase sets, each with its own stroke and height lists.

The Thanksgiving pumpkin no longer has a jack-o'-lantern face

The pumpkin picture on the free Thanksgiving tracing worksheets had two eyes and a zigzag mouth. The outline came from Tabler Icons, whose only pumpkin is a carved one. That's fine for Halloween, but it's wrong on a Thanksgiving page or in any autumn book a seller makes with the word.

Now Trace Press draws it without the face, and with two ribs, so it doesn't read as an apple. The change is made when the pictures are built from the icon set, so it can't be lost in an update. The Tabler licence (MIT) allows the change. It applies everywhere the word "pumpkin" is used: the Halloween tracing worksheets, the Thanksgiving set, and any book you type it into.

Where each page's pictures come from, and the ones with another name

The free word pages used to share one line about where their pictures come from. Now each page says it for its own twenty words. If you're selling a book made from them, this tells you which licence covers which picture.

• The animal tracing worksheets: eleven pictures from Tabler Icons (MIT licence) and nine from Lucide (ISC licence): rat, bird, worm, snail, panda, rabbit, turtle, shrimp and squirrel. Tabler has no icon for any of those nine.
• The food tracing worksheets: fifteen from Tabler and five from Lucide. Nut is one of the five, because Tabler's nut is the kind that goes on a bolt.
• The rest are all Tabler. Some words use a picture with another name, which is worth knowing before a child sees the page. On the Christmas set, ball is a basketball, reindeer is a deer's head, ornament is a bauble and gingerbread is a gingerbread man. On the picture word set, boat is a sailboat.

A test reads each page against the icon list Trace Press builds its pictures from, so if a word's picture changes, the page has to change with it. The licence texts are linked from every one of those pages.

The same twenty words, easiest first

A word's length isn't the same as how hard it is to write. Seven lowercase letters take two pencil strokes: f, i, j, k, t, x and y. So a short word can take more strokes than a longer one.

Each free word page now lists its twenty words by strokes, fewest first, and names the word that moves furthest. On the picture word tracing worksheets, fish has four letters but six strokes, so it goes from eleventh by length to last. On the Christmas set, gift goes from fourth to eleventh. On the transportation set, forklift takes 13 strokes, more than helicopter.

Each list has a link that opens Trace Press with the words already in that order, so the book gets harder page by page. A test counts the strokes from the letter shapes Trace Press draws, so the order is the book's own.

An alphabet chart, on the first page of practice

A Trace Press book can now start with an alphabet chart: one page with every letter, A to Z, capital and lowercase side by side, and the numbers 0 to 9. The letters are solid, with no dots or arrows, so a child can look at the chart while writing on the pages after it.

It follows the book's handwriting. A print book gets print letters in the same shapes as its tracing pages. A cursive book gets the chart from the free cursive alphabet chart page. It comes after the title, copyright and "This book belongs to" pages, before the pre-writing lines and A. It's one of the 52 extra pages, so the largest book is still 78 pages, under the 79 where KDP allows spine text.

A test puts it on every trim, with and without bleed, on a left and a right page, and checks that every stroke stays inside KDP's margins and no two characters touch. The book's listing description mentions it when it's on.

A free alphabet chart in print

There's now a free printable alphabet chart in print, next to the cursive alphabet chart. It's one page: Aa to Zz, four pairs to a row, then 0 to 9, on handwriting lines. It's the same page a book gets when you tick "Alphabet chart".

The letters are drawn from the same strokes as the tracing pages, so a few of them don't look like the letters in a printed book. The a and g are a circle with a line down the right, not the typed shapes. The tails tell g and q apart: g's curls left and q's curls right. Capital I has bars, but lowercase l and the number 1 are the same single line. The 4 is open at the top. The page lists these, so a parent or teacher can check them against the handwriting model their school uses before pinning it up.

If you're making a book to sell, the page has a link that opens Trace Press with the chart already ticked.

Back-cover text

The back of a Trace Press cover used to be plain. Now there's an optional "Back cover text" box under Author. Whatever you type there is centred on the back in white. A blank line starts a new paragraph.

It starts at 14pt and gets smaller if the text is long, down to 9pt at the smallest. It stays 0.375" inside the trim and stops above the lower part of the back, where KDP prints the barcode in a 2" × 1.2" box. The front of the cover doesn't move. The tracing book cover size page now says where it goes.

A test writes 1,200 characters of text, the most the box takes, on every trim and checks that every line is inside that area and that no words are lost.

See the cover while you type

The home page now draws the whole cover under the cover form: back, spine and front, the way they sit side by side in the PDF. It redraws as you type the title, subtitle, author or back-cover text, and when you change the trim size, paper or number of pages. So you can see where your back-cover words land, and how the title wraps, without downloading anything.

The preview is drawn from the same layout and the same fonts as the PDF, so the line breaks match. The free cover PDF still has PREVIEW across the front. The tracing book cover size page has the cover widths for every trim and page count.

Seven keywords for KDP

KDP's details page has seven keyword boxes. KDP's help says to "use up to seven keywords or short phrases". It asks you to leave out words already in your title, claims like "best", words like "new", quotation marks, and brands you don't own.

Trace Press now lists seven phrases under the listing description, each with its own Copy button. They come from what's in the book. A book with numbers gets "number tracing 0 to 9". A cursive book gets cursive phrases. A book with the alphabet chart gets "alphabet chart". Any phrase that's already in your title or subtitle is left out, and the next one moves up. There are no ages or claims about the child, only what's on the pages.

If you're still working out the trim size, the cover size table has the full-wrap width for every trim and page count.

The front page catches up

Today's three additions — back-cover text, the cover drawn as you type, and seven keyword phrases — were only on the tool itself. The Trace Press home page's search description, its list of what you get, and the README on the open-source mirror now name them too. So the whole job is listed: the interior, a full-wrap cover, and the listing (description and keywords).

Seven keywords on Puzzle Press too

The keyword phrases Trace Press got this afternoon are now in Puzzle Press, under the cover note, each with its own Copy button. They come from what you put in the book:

• the puzzle type ("word find book", "crosswords with clues")
• the themes you ticked: an Animals word search book gets "animals word search"
• large print
• graded difficulty ("easy to hard word search puzzles")
• a 4×4 or 6×6 grid on a sudoku book

If a phrase is already in your title or subtitle it's left out, because KDP asks you not to repeat them, and the next phrase moves up. While writing the lists I cut a few phrases that looked like other publishers' product names. KDP asks you to leave out brands you don't own.

A description for the listing, too

Under the keywords, Puzzle Press now writes the text for KDP's Description box from the book you've set up. For a crossword book it says how many puzzles and pages there are and the trim size. It also lists the numbered Across and Down clues, your themes, the difficulty, how many solutions sit on each page at the back, and the ruled Notes pages at the end. A hard word search says its words run in all eight directions, backwards too. A sudoku book says every puzzle has exactly one solution, which the generator checks for each puzzle.

There's nothing about how relaxing or fun it is. That's yours to add if you want it. It's plain text, so it pastes straight into KDP, with a Copy button.

Sofía, not Sofa

The free name tracing sheet left out any letter it had no strokes for. "Sofía" printed as "Sofa". "José" lost its é and "Zoë" lost its ë. The page said accented letters weren't drawn yet, but a name sheet that misspells the name isn't usable.

Accented letters now trace as they're spelled: á é í ó ú ý, à è ì ò ù, â ê î ô û, ä ë ï ö ü ÿ, ã õ ñ, and å, in capitals too. Each is drawn as its plain letter, then the accent as one more stroke from the top, with its own start dot. On a lowercase letter the accent sits where the dot of an i goes. On a capital it sits just above the top line, in the space between rows.

The same letters work on the cursive name sheet, in the word tracing tool (piñata, café), and on word pages inside a book. Hyphens, apostrophes, ç and ø still aren't drawn. The note under the name box says when something was left out.

Mary-Kate, O'Neil, François

The name tracing sheet dropped hyphens and apostrophes, so "Mary-Kate" printed as "MaryKate" and "O'Neil" as "ONeil". The ç was dropped too, so "François" came out as "Franois". All three trace now.

A hyphen is one stroke, left to right, across the middle of the lowercase letters. An apostrophe is one short stroke down from the top line. A curly apostrophe, which is what phones type, is made straight. The ç is a c, then a cedilla as a second stroke: it starts at the bottom of the c, drops below the line and hooks back to the left, with its own start dot.

This works on the cursive name sheet, in the word tracing tool (don't, garçon), and on word pages inside a book. ø and other punctuation aren't drawn yet, and the note under the box still says when something was left out.

A KDP cover is not 17 × 11

A KDP paperback cover is one PDF: back, spine and front side by side, with 0.125" of bleed on each outside edge. So an 8.5×11 book's cover isn't two 8.5" panels, 17" × 11". A 26-page tracing book, one page per letter, on white paper is 17.309" × 11.25". The spine adds 0.0586", which is 0.002252" a page. At 78 pages it's 17.426" wide. On cream paper, which is thicker, 26 pages come to 17.315". KDP allows spine text only from 79 pages, so these spines stay blank.

Those numbers are now a 27-second YouTube Short, scheduled for 10-06 00:30 UTC. The table with every size, for all six KDP trims from 26 to 78 pages on white or cream, is tracing book cover size. The figures come from the same code that sizes the cover Trace Press makes. Its fact-led Shorts are a test: on Buffer's counts the Puzzle Press Shorts that led with a KDP fact had 59 to 77 views, while the Trace Press tool Short had 12.

One heading per page, and it's the topic

Each page has one top-level heading, <h1>. Every page on Trace Press had the same one: the site name, "Trace Press". What a page was about, like "Free name tracing worksheet generator" or "Tracing book cover size for KDP", sat one level down, in an <h2>. So 31 different pages shared one identical top heading, and none of them said what it was about.

Now the site name in the top bar is plain text styled exactly as before, and each page's own title is its <h1>: for example name tracing, cursive letter tracing and the KDP cover sizes. Nothing looks different. Before and after, I measured each heading's font, colour and position and each page's height on all 31 pages, at desktop and phone width, and all of it matched.

A test now checks that every new page has one <h1>, that it isn't the site name, and that it shares a word with the page title.