Adding these lines to my application.env covered this issue

# do not fetch external resources
DISABLE_FETCH_PREVIEWS=true
DISABLE_EMBED_PREVIEW=true

I will need to go through my own feed next to ensure that there are no embeds left over. This is in addition to my recent purging of boosts: https://mastodon.mariobreskic.de/@mario/114982417840616548

#server #asca #mastodon #gdpr

And from there, I delete all #PreviewCards in #mastodon by using these commands:

1. docker compose exec web bash
2. RAILS_ENV=production bundle exec rails console
And then I run this in the console: PreviewCard.destroy_all

This will run for a while but it should remove all stored preview cards from my mastodon server, for all posts.
And it worked. Just climb back up to my prompt by punching in “exit” until I am back home, and that issue has been fixed as well. #server #asca #mastodon #gdpr

Correction, see #aside here https://mastodon.mariobreskic.de/@mario/115050605879471525

#aside and #tldr
Running a solo Mastodon instance? Here's how to keep it clean and #gdpr compliant:
– Disable link previews: DISABLE_FETCH_PREVIEWS=true
– Purge old PreviewCards via Rails console
– Avoid federated content
– Self-author everything
– Host locally, anonymize IPs
Respectful publishing, no third-party hosting
#asca #server #mastodon #selfhosted #germany

There seems to be a bug implementing this in #mastodon #v4.4.3, so here is the workaround for disabling fetch previews by deleting preview_cards from the DB:

DELETE FROM preview_cards;

And blocking future inserts:

CREATE OR REPLACE FUNCTION block_preview_card_insert()
RETURNS trigger AS $$
BEGIN
-- Ignore insert
RETURN NULL;
END;
$$ LANGUAGE plpgsql;

CREATE TRIGGER prevent_preview_card_insert
BEFORE INSERT ON preview_cards
FOR EACH ROW
EXECUTE FUNCTION block_preview_card_insert();

Undoing the previous changes inside the db container, due to having found a more promisingly elegant solution:

DROP TRIGGER IF EXISTS prevent_preview_card_insert ON preview_cards;

DROP FUNCTION IF EXISTS block_preview_card_insert();

Fix:
adding this to my yaml file for my sidekiq and my web container:

command: >
bash -c "bundle exec rails runner 'class FetchLinkCardService; def call(*args); nil; end; end' &&
bundle exec puma -C config/puma.rb"

and sidekiq

command: >
bash -c "bundle exec rails runner 'class FetchLinkCardService; def call(*args); nil; end; end' &&
bundle exec sidekiq"

Filed a bug: https://github.com/mastodon/mastodon/issues/35813

Which is also a test for the patch ;)

#server #asca #mastodon #gdpr #PreviewCards

I need another test, as a sanity check:

https://de.wikipedia.org/

#PreviewCards are still being created, despite the patches.

Changed the #sidekiq patch to target the LinkCrawlWorker like so:

command: >
bash -c "bundle exec rails runner 'class FetchLinkCardService; def call(*args); nil; end; end' &&
bundle exec sidekiq"

And another one:

http://web.archive.org/web/20140903082741/https://archive.wired.com/techbiz/people/magazine/15-10/st_thompson

And brute force I will now, since the Worker insists on working.

Inside the pgsql container:

CREATE OR REPLACE FUNCTION block_preview_card_insert()
RETURNS trigger AS $$
BEGIN

RAISE NOTICE 'PreviewCard insert blocked by DB trigger.';
RETURN NULL;
END;
$$ LANGUAGE plpgsql;

CREATE TRIGGER prevent_preview_card_insert
BEFORE INSERT ON preview_cards
FOR EACH ROW
EXECUTE FUNCTION block_preview_card_insert();

And another test, which I hope completes this issue for me:

https://www.schulmanart.com/2017/03/how-to-price-art/

Rebuilt the index post/persistent context “tool” for tumblr because threading there should not be done with reblogs.
Instead, I used a system I came up with in 2019, on Twitter: make a post containing a link I want to hook up to the persistent context scaffolding.

On tumblr, this means: make a post, then link to that post from its category/theme master post.
https://www.tumblr.com/codeandcanvas/792786451752468480/index-post #asca

Will have to update my #linkwarden incrementally from v2.11.5 to v2.12.2.

I use linkwarden to store links from my blog articles via a custom wordpress plugin which I have hacked together using #chatgpt #asca

I am sure that doing the update incrementally makes more sense than jumping a few versions in between.

Restarted my side project of bringing the @medienfeed service to the fediverse.

A free info service about design and media for the #DACH region, which already runs well on other websites.

Played around with making #huginn act like a buffer for the posts, using a Scheduler Agent and a Delay Agent for the scheduling.

Added a truncate filter to the Formatting Agent so the process doesn’t get denied by exceeding xxx amount of characters.

Unter uns, was sich Firmen in puncto Nutzern einbilden, ist entsetzlich.
ChatGPT meinte neulich dazu:

‚If you participate meaningfully, you are infrastructure.
And infrastructure doesn’t get to say “nah.”‘

Ich finde die Grundlage des Lagrange-Punktes 4 dabei reizvoll: nicht nur handelt es sich hierbei um Objekte, die unter einem viel stärkeren Einfluss von Massen stehen, die ich selbst nicht beeinflussen kann

(Social-Media-Websites sind im Allgemeinen nicht stabil, die Accounts, die man dort hat, können durch Trends, Performanz und Cliquen sehr schnell verschwinden, dem Tribalismus ist das zu schulden),

man lebt auch schlicht nicht auf Trojanern. Aber man kann Sonden dort absetzen. Das geht.

Ich überarbeite mein Modell: Social Media sind nicht L4 oder L5, sondern astronomische Zentauren.

Instabil.

Versuche gerade meine Bookmarks von wallabag zu Linkwarden zu migrieren. Die Frage ist, ob Datengräber wie Linksammlungen nicht doch einfach nur nach einer Variante der Pareto-Regel für Daten funktionieren: 80% des Speicherplatzes sind immer von den Sachen belegt, die man nur in 20% der Fälle benötigt.

Aber gut. Irgendwo steht eine Notiz, dass der Bibliothekar (und auch der sogenannte „Hoarder“, also ein kompulsiver Sammler) im Mensch 2026 sublimiert wurden.