Acorn DevOps
@acorndevops.bsky.social
102 followers 302 following 2.5K posts
acorninteractive.ca Home of the devvy Kjörndogs. Buckle up, get comfortable. We’ll be pushing automated updates from any major developments on our core application, while publishing technology insights, code related findings, and general dev musings.
![ASCII flow diagram on a dark background mapping the blog workflow from Google Doc to live post, part 1 of 5. Step 1, “GOOGLE DOCS — authoring + editorial pass [OUTSIDE APP]”: the writer drafts, the editor comments inline, then sign-off happens in the doc; it lives entirely outside the app with no record in Mongo. The arrow down notes pasting HTML into Slate nodes works (withPasteFormatting.ts), but importing via a doc link or the Drive API is a gap, with no ingestion code at all. Step 2, “CREATE POST SHELL [OK]”: SeoModal goes through dispatch.ts to blog.create, setting a unique kebab-case slug, title, description and categories, with status draft, no publish date, and the caller as author. Step 3, “ASSIGN AUTHOR (optional — guest contributors) [OK / ~]”: AssignArticleModal or GuestContributorsPage calls blog.assignAuthor, writing the user ID, assigned-at and assigned-by; the role check is a hardcoded string compare rather than the RBAC helpers.](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:n3xrtzprp3blxeqiid5abew3/bafkreidfavmcnncw3eljiuhlinaebdb75r4xgqhdu3i7dxlidzjlsqppi4)
![Part 2 of the diagram. Step 4, “COMPOSE BODY — dynamic templates + components [OK]”: BlogTemplate feeds BlogDispatcher, a picker that reads the component list, then SectionSelector switches on a key to render a block component. It lists 22 block keys, including text editors, images, gallery, video player, tweet and Bluesky embeds, contact forms, share card, code block, four circle and square layouts, sequence blocks, staff quote, staff grid and we_nuts. The Slate text editor autosaves through blog.edit. Step 5, “REVIEW / APPROVAL [MOSTLY GAP]”: the “Submit for Approval” button opens BlonkchainsModal, a joke modal, with no server call, status change, notification or email. The status options are only draft, live or archived, with no review state and no reviewed-by, approved-by or submitted-at fields. The coordinator’s “Awaiting Approval” dashboard queue is just anything not live, so approval actually happens back in Google Docs or Slack.](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:n3xrtzprp3blxeqiid5abew3/bafkreiglu2tx4ttoo4yfeif5eam5i4mzxzns776krgdwy26jlsqsvjiheq)
![Part 3 of the diagram. Step 6, “PUBLISH GATE [OK / GAP]”: the client refuses going Live without a listing image (OK), but the server’s blog.toggleStatus only checks that a user is logged in (GAP); the BLOG_PUBLISH permission is defined but never called, so any logged-in user, including a guest contributor, can go live. Step 7, “GO LIVE [OK + GAP]”: toggling to live sets the status, sets the publish date on the first transition only, and appends a status history entry; there is no cron or scheduler, so future-dated publishing is impossible, and a Zapier webhook sits there commented out with a TODO to uncomment. Step 8, “SERVE [OK]”: listing and single views filter to live posts, server-rendered through BlogProvider and BlogTemplate with a 5-minute client cache, and SeoTags emits one social share image into Open Graph and Twitter tags, so every network scrapes the same title, description and image.](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:n3xrtzprp3blxeqiid5abew3/bafkreielpoyl5pto3khrhfdwxmgk4pw3iokhjiauerncweqspxqhiqszla)
![Part 4 of the diagram. Step 9, “SYNDICATION — per-provider tailored captions [NOT BUILT]”: today, share icons are reader-initiated click-to-share only, with hardcoded copy for Facebook, X and LinkedIn, no Bluesky or Instagram button, and nothing posted outbound. A proposed data shape stores per-provider caption, image, hashtags, link mode, first comment, status (queued, posted, failed or skipped), posted time, remote ID and URL, and who posted it. Next, “CAPTION COMPOSER” with one tab per provider and a live character budget, plus fallback chains so nobody has to fill all five: caption falls back from provider caption to default caption, excerpt, then description; image falls back from provider image to the social share image, then a default. A guard refuses going live if any enabled provider’s caption is over budget. It fans out to five provider boxes: Bluesky, 300 characters, links OK, a question about alt text; X, 280 characters, links OK, thread split; LinkedIn, 3,000 characters, links OK but “hates links”; Facebook, long OK, links OK, Open Graph image rules; Instagram, 2,200 characters, no links, link in bio.](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:n3xrtzprp3blxeqiid5abew3/bafkreidamxzdj7dtfkf2kath6vdt7rnnbrzg4qgef4i6e2zedsqfeqjjqm)
![Screenshot of Meteor documentation in dark mode, explaining the Email.sendAsync method. The text reads: "sendAsync only works on the server. It has the same behavior as Email.send, but returns a Promise. If you defined Email.customTransport, the sendAsync method returns the return value from the customTransport method or a Promise, if this method is async." Below is a JavaScript code example. It starts with the comment "Server: Define a method that the client can call." It calls Meteor.methods to define a method called sendEmail that takes to, from, subject and text. Inside, a comment reads "Make sure that all arguments are strings," followed by check([to, from, subject, text], [String]). The method then returns Email.sendAsync({ to, from, subject, text }) with a .catch((err) => { }) handler containing the comment "do something with the error."](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:n3xrtzprp3blxeqiid5abew3/bafkreihb2vjmewgrke7o6iwlub2efv4aet3tyzez37w5zgtuibtykllfyq)
![Text slide. Top annotation "WHY TWO LANES, NOT ONE": unit alone → rules provable but "does login actually work?" stays unknown; e2e alone → flow provable but slow, flaky, and can't enumerate every role; together → each role asserted cheaply in unit AND end-to-end in e2e. Below, a Footnotes section — "how this builds off types & existing coverage toward auth integrations" — with three notes. [1] Rides the type declarations as the single source of truth: both lanes assert against the same role enum, User type, field-projection constants and input schemas, so a type change surfaces as a failing test, not silent drift. [2] Extends the existing test harness instead of forking one: reuses the mock method registry, in-memory collection, and accounts mock; existing patterns were the template. [3] Positions a verified base for auth-based integrations: future work — attempt/anomaly logging via runtime hooks, a security harness, or outward auth (SSO, API-token, third-party identity) — inherits the same two-lane contract: a unit proof of the rule plus an e2e proof of the flow.](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:n3xrtzprp3blxeqiid5abew3/bafkreiftuzdb64vx2fa7zwqtxdkduyxmeq74fgp7uc2bhfa4xszf7k4bqu)
