Tracking and advertising are orthogonal, you can have the latter without the former. The main challenge is that context-based ads (as opposed to tracking-based ads) don't work very well for websites that don't have much relevant context, such as news sites. I'd argue that news sites should either be subscription-based or public funded. The latter already happens to some extend in the EU.
Automated systems are wonderful. A few days ago I called off a dentist appointment for today, moving it to next week instead.
Now I just got a message that the dentist happens to have a spot free today, perhaps I'm free as well and would like to move next week's appointment back to today?
I have strong opinions about certain topics but I'm often not able articulate those opinions very well. Maybe I should try writing more. :blobcatthinking:
Q: The systemd journal manages logs in a binary format with metadata for efficient filtering. What is the best way to query the postfix logs in a default setup on Debian Bookworm?
Revamping my email setup again, this time with Postfix.
...which truly is software from a different time. Surely, even 25 years ago, user@hostname rarely referred to an actual system user at the given hostname.
@lanodan That sure would be funny, but I expect the Linux VFS layer to support longer names; most(?) syscall arguments are limited to PATH_MAX rather than NAME_MAX. Need to give it a try, though. :blobcatthinking:
@lanodan Oh yeah, I forgot to document that ncdu doesn't support indefinite-length strings for the name field, they must be length-prefixed (That is such a CBOR misfeature).
The maximum length is actually a bit larger than that; the root item contains the full path of the scanned directory instead of just the name, and some filesystems apparently allow for 255 codepoints instead of bytes: https://en.wikipedia.org/wiki/Comparison_of_file_systems#Limits
@lanodan Working on a new export format that supports parallel export and random-access browsing without loading the entire directory tree into memory. Could work without compression, but that's somewhat more of a waste of disk space than a larger binary.
The good news, compiling libzstd with ZSTD_LIB_MINIFY=1 brings it down to 164k while still easily beating zlib (100k) in terms of performance and ratio.
The bad news, 164k is still pretty significant on the ncdu static binary:
Full-time unemployed as free software developer and sysadmin. Part-time dog parent, bookworm, gamer, weeb and minimalist.Posts are deleted after 1 year. #nobot