For a sneak preview of my followup to the Fediverse Schema Observatory, there's now a project page on the lab website for the ActivityPub Fuzzer:

asml.cyber.harvard.edu/activit

Basically it lets an ActivityPub server developer test compatibility with simulated messages from Fediverse software known to the Observatory (dozens of software projects and hundreds of versions!). The dev can do this entirely in a local dev environment.

The project is still in the works but I'll be releasing the code soon.

@darius

super nice.
#fedidev#ActivityPub

What I worked on the past weeks was to analyze the current news corpora of the University of Leipzig in different languages.
Then mixed it with an anonymized fedi corpus and wrote a thing which can
- compress ActivityPub Objects to 20% of its size by a combination of semantic compression where 256 languages can be covered. The rest would be uncompressed in multilanguage Objects (see Evans new Primer Page).
Result is UInt8Array for a database.

What I am working on now is to "preserve hashtags and common knowledge".
One byte is a pointer to Hashtags (where any word has a # at 0) and one byte is a pointer to 3x(256²) wikidata tables. So that we directly get the e.g. Q1055 for Hamburg and can ask the author if the prominent Hamburg is meant …
This was just a first demo for German https://github.com/sebilasse/compressDE which meanwhile improved and soon you can generate your lexica from corpora.
btw This can detect 852 languages https://github.com/redaktor/languages
Less data, better climate …

@darius I meant the post here, which presumably can reach more of us who are non-devs.

Like, I couldn't tell if the post here was about preventing or detecting when someone outside fedi is exploiting a code vulnerability that allows them to access fedi in what is otherwise a legit way, or if it was about preventing or detecting when someone outside fedi is misusing the code by some other method, or, you know, something else.

I hope that helps, thx.