Be sure to put your pronouns in your bio (or display name, if you want)
Even if you're not someone who disagrees with what the pronouns you were assigned at birth, it helps normalize sharing our pronouns and helps support those who do.
Element (Matrix chat app) suspended from the Google Play Store - https://news.ycombinator.com/item?id=25964226
I am now evolving into a Decentralised Libre Zealot - Join me.
A Jenkins job fetches upstream every day for each repo: https://jenkins.marussy.com/view/mirrors-and-patches/ It's a tiny bit smarter than `git clone --mirror`, because it hopefully detects forced updates from upstream and tags them.
Should be pretty DMCA-proof, because the server is not in the US.
Also, my own branches can also live it the repos this way. My customizations get automatically merged after every pull, so it's CI for my patches (that I don't necessarily want to submit to upstream).
Here's the script: https://git.marussy.com/gitmirror/tree/ *adds it to the pile of stuff that need documenting*
The other way to do it is to puppet Twitter accounts of everyone who replies, but then 1. only people with Twitter accounts get their replies bridged back to Twitter 2. why bother with following people through a Twitter bridge it you're okay with having a twitter account.
Not to mention that even the Twitter->Fedi direction could be a bit annoying. You'd have to make sure the bridge never federates with a Fedi relay, otherwise you're likely piping the Twitter "firehose" into small Masoton/Pleroma instances that are unprepared to handle the increased load. Following bridged users directly would be fine, though.
But honestly, all of them feel slightly shady (it's impossible to prove in a centralized service that they aren't gathering data), although not to the extent Google does. Was thinking about setting up a #Searx instance, but all the (proxied) searches coming from my server's IP would be a huge fingerprinting vector.
The approach targeted by #Dendrite -- a bunch of small (even single-user), federated servers -- seems to me as the best solution for communication if full peer-to-peer turns out to be infeasible (looks like there must be some kind of server for offline messaging, unless DHT can solve that problem somehow...)
Some random lessons learned after starting to use Beancount:
1. I find accounting on an accrual basis unreasonably fun, and that's somewhat worrying.
2. Lacking a standard data format, parsing statements for financial institutions is quite finicky. Sorry, no tests in the repo, as I can only use my actual live data as tests.
3. The uni has properly paid all my salary in 2020. The schedule of bank transfers was a bit idiosyncratic, though, but in the end it adds up correctly despite the uni adopting SAP this year and other opportunities to screw up bookkeeping.
4. By parsing all my debit card transactions, I did notice a majority of eating at fast food places around the uni instead of healthier stuff (that is, during months we weren't WFH). I should probably work on changing my habits for the better.
5. Type annotations make Python an adequate language to code in. However, the situation with Python build tools is still atrocious.
Just tried both Poetry and Pipenv, but they are insane and deep-clone any Git dependencies from scratch whenever any package is updated...
Currently it just mirrors some software from Github I contributed to or use, but I'll put my own code here also.
Still to go: try to set up Prosody for XMPP, Dendrite for Matrix, and maybe Jitsi 🤓
There's a culture of tech that is about humans interacting with machines. As if it's only a logical/technical activity. No room for emotions and all that soft stuff.
Then another culture of "tech" that's really about rich people extracting wealth (via data) from other people. Harnessing tech for some goal.
I want the culture of tech, where it's about people interacting with people, with an unobtrusive technology layer to facilitate an essentially human activity. Tech as a supporting tool.