Release the kraken
9 minperlhugoindexnoxautomationdeploymentnetlifyseoblogreleasepublication

TL;DR

Netlify’s build-on-push burned through my tokens fast. So I built Blog::Release — a Perl module that handles scheduling, deployment, validation, and search engine pings on my terms.
The idea is simple: A configuration declares the state and I write, queue it up, and the machine handles the rest. Holiday support is baked in because even if the backlog is stacked, we aren’t animals. Content validation keeps me from publishing embarrassing SEO disasters. And when something goes out, search engines get poked automatically. Except Google, who can get fucked.

I don’t want to think about when to publish. I want to think about what to write.

Introduction

A while ago when I wrote about my mixtape I rediscovered blogging. And because I wanted to get rid of dev.to (and Medium before that) I moved to my own blog with Hugo. Why should big tech feed of my content? They don’t give me reader insights, they just give you a platform where you can type text and add images. You can do that without sponsoring their bottom line. And btw, even now I don’t have reader insight, so say “Hi” if you read this. I digress :)

So I took a hugo template and republished my blog. But I had a problem. When you set up Netlify they do something clever, they build on push events. That triggers a ton of rebuilds everytime you change a little item in your blog and you push the change. I was unaware of the token aspects of Netlify, so… within hours I hit 80% of used tokens. For what? Committing early and often (and pushing it).

Blog::Release

So I decided to build a release pipeline for this blog. And before I did that, I had some thoughts on how to approach blogging. I wanted to write content and publish it on a given day to my liking and to have some kind of schedule. I didn’t want to manually release blog content. Not on the regular at least.

I built an initial version of what now is Blog::Release. It uses WebService::Netlify under the hood to talk to Netlify and do the deployments and everything is coordinated from a Makefile:

zsh 
$ make
Available targets:
  make configure       - Install Go and Hugo on Debian (may need: sudo make configure)
  make dev             - Run Hugo development server natively
  make build           - Build site natively
  make clean           - Clean build artifacts
  make realclean       - Clean build artifacts including resources
  make publish         - Publish the hugo site to Netlify
  make deploy          - Deploy (no publication changes) the hugo site to Netlify

Available docker targets:
  make docker-build    - Build Docker image
  make docker-dev      - Run Hugo dev server in Docker
  make docker-dev-live - Run Hugo dev server in Docker
  make docker-clean    - Remove Docker image
  make docker-install  - Install dependencies inside Docker (used by Dockerfile)

I wrote a little wrapper and.. the site deploys:

zsh 
$ crontab -l
0 5 * * * $HOME/wesley.schwengle.net/bin/deploy.sh

Scheduling your articles

This is the actual cool bit of Blog::Release. This is where most of the logic is hidden. I have a yaml file that essentially says publish these articles in this cadence. The order matters: FIFO. Well, you can change the FI, thus the FO too.

yaml 
publish:
  - name: your-article.md
    type: article
    # optional fields
    date: YYYYMMDD
    finished: false
    announce: |
    Announcements go here
  - name: ...

This tells us a couple of things. One, what the article is, where to find it, and when it should be scheduled. The scheduling is dynamic. You can either opt for a manual date or you opt for a automatically scheduled date. The cadence is defined in days from a given moment. Why? Well, I didn’t want to schedule a once a month kinda thing. I just looked at a calendar and said: every week on Tuesday I want to schedule my articles. Or every three weeks. Or every day. It may limit you in every Tuesday or Wednesday, but.. I ain’t blogging that much to support these corporate style release trains. Plus solely based on days I can do cool tricks without having to do too much math with dates.

yaml 
release:
  # Push updates in a release bundle (false) or on separate days (true)
  fixAlways: false

  # Every 14 days on a tuesday from 2026 onwards
  startDate: 20260113
  interval: 14

  includeDate:
    - 20261208
  blockDate:
    - 20261229

  # Yell at us for keeping more than 4 published articles in the queue: clean
  # your shit up
  maxPublished: 4
  # Yell at us for keeping less than 4 published articles in the queue: write
  # more or just plug them in
  minPublished: 4

  # Support holiday support, we aren't animals, we gentle people
  holidays:
    - AW
    - NL
    - BQ
    - CW

You’ll notice some things that I haven’t explained yet. fixAlways and holidays I’ll do first.

When a blog gets published Blog::Release tags the release with YYYYMMDDXX tags. I stole that from old skool DNS serials. Easy on the eyes and clear in meaning: 00, 01, 02, you know exactly which tags were released and Netlify also gets them.
Now, when you have published an article in the past and you have edited them Blog::Release sees this and schedules it for publication. You can set fixAlways to true it will tweak some metadata so when hugo comes along it updates the lastmod and your content got refreshed. When you set it to false it will only schedule them for publication when you have a manual or automatic deploy happening. So it gets bundled to your normal release cadence. It is what I prefer. I can make changes and know nothing will change under my nose except when a publication is actually happening.

Holiday support is baked in because even if we have a backlog, we want to appear mindful to our readers. They are not going to read your blog on Boxing day or when they celebrate the flag or whatever your local holiday schedule day says it is. You can even use your own personal holiday unit, so it doesn’t publish on your birthday. Date::Holidays says hello. For myself I support the Aruba, Bonaire, Curaçao, and the Netherlands because those are my roots and I live in the area.

The includeDate and blockDate configuration allows you to punch holes in your queue or to block them. They still adhere to holiday rules. I added them because in my case Dec 15th is a holiday and Dec 29th is a non-holiday release slot. Yet, I don’t want to publish between Christmas and New Years. So I block the 29th and I open up a hole on the 8th, which is normally not a release date. If you really want to publish on a holiday, you’d have to schedule it manually. Holidays are treated as “You want THIS content to go out today”, not via the queue.

The minPublished and maxPublished are for validation purposes. The scheduler doesn’t remove entries once things got deployed because I like comments in my configuration files and most parsers strip them out when editing, so I baked in a git pre-push hook to ensure I get warnings once these thresholds are violated. maxPublished is for when you have more than X published items sitting in the publish queue. And minPublished is for when you have less than X articles scheduled for publication.

Validation

There is also content validation built in, I had it mainly to support SEO things, but truth be told, my own SEO sucks. But it’s there and I use it. There are some hardcoded rules: You must have 2-10 tags, 10-20 keywords, and a description. The slug gets generated on publication, so that isn’t checked.

zsh 
$ blog-release validate

Queue status:
  ✓ 0 already published (max: 4)
  ✓ 19 unpublished scheduled (min: 4)

$ blog-release validate --strict

[snip]

» content/article/blog-release.md
  Warnings:
  ⚠ Description length 0 outside range 100-180 chars
  ⚠ Need at least 2 tags, found 0
  ⚠ Need at least 10 keywords, found 0

[snip]

Tags and keywords

In order to automate a bit of tags and keywords for “Related content” usage, I dove into NLP things and more specifically, how to extract keywords from text and how to score that. So Lingua::TermWeight was born to support that tag and keyword generation. This added a Lingua plugin/command: blog-release lingua.

You can configure how tags and keywords are used:

lingua:
  plugins:
    - Blog::Release::Shortcode::Details
    - Blog::Release::Shortcode::Links
    - Blog::Release::Shortcode::Color
  stopwords:
    - didn
    - please
    - get
    - ...

You can write your own plugins to strip so-called shortcodes or other items in your articles and you can add stopwords to ensure tags and keywords don’t contain words you don’t want to have. And in addition, if you have keywords and tags manually set. Lingua will fill the remaining with the units found from your blogpost.

Publication

There are several publication modes

text 
⚙  auto       normal scheduled publish
󰢧  manual     manually scheduled posts
↻  republish  fixes/updates to existing posts
󱦹  draft      not part of the release system yet
⬈  published  already released

When you run blog-release schedule --show it will show you output like this:

text 
20260324:
  ⚙ noguiboot.md
  ├ Question: My Debian doesn't boot with a GUI
  ├ Troubleshoot Debian not booting to GUI by checking display manager statu...
  └ content/answerrit/noguiboot.md

  ↻ 20240505-git-fetch-tags (bundle)
  ├ Limiting tags when using git fetch
  ├ Use negative refspecs and tagOpt to filter out CI-generated tags when fe...
  └ content/article/20240505-git-fetch-tags/index.md

  ↻ 20250823-mixtape (bundle)
  ├ Mixtaping in 2025 – A Personal Build Log
  ├ Building an offline HTML mixtape with web components, custom video/audio...
  └ content/article/20250823-mixtape/index.md

  ↻ 20250924-workspace-on-demand.md
  ├ Workspace on Demand in i3wm
  ├ Build dynamic context-aware workspace groups in i3wm using Perl, AnyEven...
  └ content/article/20250924-workspace-on-demand.md

20260407:
  ⚙ docker-cpan-aua.md
  ├ Docker to the rescue
  ├ run perl cpan modules in docker containers with minicpan and cpm to avoi...
  └ content/article/docker-cpan-aua.md

20260421:
  ⚙ update-alternatives.md
  ├ Question: How do I register deepin-terminal as an alternative on Debian?
  ├ Register custom terminal emulators in Debian's alternatives system using...
  └ content/answerrit/update-alternatives.md

And draft, like this post at the time of writing, are shown like this:

text 
20261020:
  󱦹 blog-release.md
  ├ Release the kraken
  ├ ...
  └ content/article/blog-release.md

Search engine support

Via Bing.com I discovered IndexNow. It essentially helps you poke search engines to tell them you’ve updated your site. Google for incomprehensible reasons doesn’t support it but Google refuses to index this blog so.. Fuck ’m.

The configuration is pretty straight forward, define a key, define the search engines you want to target and poke them on release. There is some nice things going on, as Blog::Release parses your sitemaps and can tell when URL’s were added, updated or removed. It will poke IndexNow on any of these situations.

yaml 
site:
  hostname: your.blog.com
  name: netlify-username
  id: netlify-site-id
  indexnow: index-now-id
  searchengines:
    - api.indexnow.org
    - www.bing.com
    - indexnow.amazonbot.amazon
    - yandex.com
    - indexnow.yep.com
    - searchadvisor.naver.com
    - search.seznam.cz

Release the Kraken!

This blogpost, and any other blogpost of mine, are published without human involvement. Well, I write the posts, but the actual publication step I don’t do. Automated via cron: blog-release publish. I just write and edit some yaml file.