12 Mar 2026
When did this file last change in Git?
Some of the posts on here get updated. For example, I change effective privacy tips as the available services and settings change. (If you’re in California, go do DROP which I added recently.) And sometimes I add a “More” link to the end of a post I wrote more about later, or just fix an error.
Previously every post just had the blog date, from the
yaml_metadata_block at the top of the source file. (more
info: Metadata
blocks) But that’s the date I originally wrote something, so it’s
the wrong date to use for an XML sitemap or for the
RSS feed of recently changed
pages. Since I build sitemaps and RSS from the generated HTML, I
need to put two dates in each HTML page. The meta tags for that look
something like this.
<meta content="2024-01-01"
name="dcterms.date">
<meta content="2024-01-01"
property="article:published_time">
<meta content="2026-02-27"
property="article:modified_time">The dcterms.date
and article:published_time
are the original date from the metadata block, in YYYY-MM-DD format. The
modified_time has to be obtained from git, using
git log, like this:
That’s gitdate.sh that runs in order to populate the
Pandoc command line. I’m using UTC here because the server is on UTC and
it would be confusing for a crawler to have HTTP
Date headers in one time zone and meta tags populated
from a different one. Besides, I’m from Indiana
originally so time zones make my brain hurt.
Bonus links
The
MacBook Neo By John Gruber. 8 GB of RAM is not a lot….But just
using the Neo, without any consideration that it’s memory limited, I
haven’t noticed a single hitch. I’m not quitting apps I otherwise
wouldn’t quit, or closing Safari tabs I wouldn’t otherwise close.
(He did say Safari tabs, not certain other browsers. But still,
I’m saving this in case I need an argument against adding complicated
crap to a site. Have we tried that on an 8GB laptop with other stuff
running?)
Grammarly
Is Facing a Class Action Lawsuit Over Its AI ‘Expert Review’ Feature
by Miles Klee. Contrary to the apparent belief of some tech
companies, it is unlawful to appropriate peoples’ names and identities
for commercial purposes, whether those people are famous or not.
Europeans are dangerously reliant on US tech. Now is a good time to build our own by Johnny Ryan. (If you’re waiting for the government to come around and tell you to protect yourself, I don’t know if even the biggest, most ambitious government program can help you. Ask your grandparents or great-grandparents: would they wait for the government to send someone to tell them to turn the lights off during an air raid? Or do you check your own lights and help your neighbors if needed? This is up to every IT decision maker, at every level, from corporate and public sector CIOs, to development project leaders. For “Europe must” it makes sense to substitute “An IT decision maker who wants to protect their users must”. Anyone making an IT choice can either accept the manageable risks of filling in technical gaps in open source or responsibly hosted software—or choose some solution that might look more feature-complete today, but holds unpredictable risks when it starts working against you. More: you can’t spell #Eurostack without U)
A
Take-Down of the Take-Down by Brian Shearer. Banks’ average
Return on Assets (ROA) across all activities is consistently around
1.1%. But the ROA for credit cards is 6.24%, reaching as high as 10.82%
for subprime customers, but still a high 2.56% for even super-prime
customers. Investors are consistently happy with 1.1% returns from
banks, so the credit card profits are much higher than they need to
be.