Source of the openbasalt.org website.
Plain static HTML and CSS, no build step. The only scripts are the analytics
tag and its small bot filter, and the gallery viewer, all described below. Served by GitHub Pages from the
main branch root, with the custom domain set in CNAME.
index.html home page
projects/index.html projects: Samba Conductor, cotty, Basalt OS and obpkg.org
projects/samba-conductor/index.html Samba Conductor: features, screens, status, install, repositories
projects/cotty/index.html cotty: what it does, screens, status, install, documentation
gallery/index.html gallery: Samba Conductor screenshots and a pointer to the Basalt OS gallery
404.html not found page
assets/site.css styles (light and dark via prefers-color-scheme)
assets/umami-filter.js keeps automated browsers out of the visit count
assets/gallery.js gallery viewer and the optional Videos section (same file as on basalt-os.org)
assets/gallery/ gallery images: <id>-720.webp and .jpg thumbnails, <id>-1600.webp
assets/projects/ images used only by a project page (same three sizes), and its og.jpg
assets/og/ Open Graph images, 1200x630, one per page
assets/ favicons and the social preview image
robots.txt, sitemap.xml
To preview locally, serve the folder with any static server, for example
python3 -m http.server 8000, and open http://localhost:8000.
Every indexable page has its own title and description, a canonical URL, Open
Graph and Twitter tags with a 1200x630 image (assets/og/, or
assets/projects/<name>/og.jpg), and is listed in sitemap.xml with its
lastmod. The home page carries JSON-LD for the Organization, and a project
page a SoftwareApplication. The 404 page is noindex and stays out of the
sitemap. When a page is added, add it to sitemap.xml; when assets/site.css
changes, bump its ?v= query in every HTML file.
gallery/index.html works like the Basalt OS gallery (see the README of
basalt-os/basalt-os.org): each image is a figure in a ul.shots with a
one-line caption and its source and date, in three files under
assets/gallery/ (720 px WebP and JPEG thumbnails, a 1600 px WebP for the
viewer). Only lab captures with made-up names, no real accounts, e-mail
addresses, internal host names or addresses, passwords or keys. Without
JavaScript every thumbnail is a link to the large image.
Terminal screens, such as the cotty ones in assets/projects/cotty/, come
from a real session in a private tmux server: tmux capture-pane -e of the
pane, drawn cell by cell to PNG with its colours, then sized like the rest.
The 720 px thumbnails of wide terminals are the left part of the capture at
16:10, so the gallery crop stays readable; the 1600 px file is the whole
capture. Use a throwaway directory, a plain prompt and a clean environment,
so no home paths, host names or personal data end up in the image.
Videos: the Videos section stays hidden until the JSON block
#gallery-videos in gallery/index.html lists at least one video, as
[{"id": "<YouTube video id>", "title": "...", "caption": "..."}]. Nothing
is requested from YouTube until a person presses Play; then the player loads
from youtube-nocookie.com.
The footer links to the project on GitHub, X and YouTube with plain links and inline icons, no widgets.
Every page loads Umami, self-hosted at
analytics.openbasalt.org, to count visits. It sets no cookies, stores no
personal data and keeps no cross-site identifier, so there is no consent
banner. The tag carries data-do-not-track="true", so browsers that send Do
Not Track are not counted at all, and data-domains limits it to
openbasalt.org and www.openbasalt.org, so local previews and forks never
report anything. The footer of the home page says the same in one line.
The tag also carries data-before-send="umamiBeforeSend", a function defined
in assets/umami-filter.js and loaded just before it. It keeps automated
browsers out of the count without blocking them: crawlers, previews and test
tools can fetch every page as usual, they are just not counted. Umami calls the
function before each report, and it skips the report when navigator.webdriver
is true, when the user agent carries a common headless or bot marker
(HeadlessChrome, PhantomJS, bot, crawler, spider, slurp, facebookexternalhit,
preview and similar, case insensitive), or when the screen is exactly 800x600
and the browser reports no languages at all. Every other report goes out
unchanged. It stores nothing and sets no cookie. When the filter changes, bump
the ?v= query on its tag in every HTML file.
Published text follows the project style: no em or en dashes, no bold, no ellipsis.