Yes, we are getting there. I think compiler is a bigger problem than unit tests given most verticals don't even have that. With unit tests, there would be some reward hacking but would be controlled at the model level + tests. (this is one of the reason i dont believe in transformer based llm as a judge for a verifier)
I think that superkuh's point is that it is not a systemd thing. Cleaning up /tmp by deleting old files has been around since before systemd was invented. Since before Linux was invented, even.
Outwith the world where only Debian and other Linux distributions exist, though, this isn't really systemd doing bad things to poor old Debian. To a wider world it is Debian and other Linux distributions exceedingly slowly reinventing things from Unix.
AIX had /usr/sbin/skulker from at least as early as version 1.2.
For AT&T System 5, Fielder and Hunter were popularizing a similar utility named rmtrash that is run nightly.
Fielder and Hunter exemplify traditional Unix thinking in their 1986 book on Unix system administration, which includes rmtrash:
> By putting all temporary files in one or two directories, it is easy to clean them all out at regular intervals. For this reason, it's a good idea to encourage users to use /tmp for all files they need only a short while.
skulker is not in the old comp.unix.aix Usenet FAQ document, but given the number of times "That's what skulker does, and you should be running it daily." seems to have been the answer over the years, it probably should have been. (-:
To old hands, this is not a systemd novelty at all. The novelty, to old hands, is the idea of doing this not using a script. I remember the war stories from the 1990s and turn of the 21st century when users did things like include LF in the names of temporary files, and administrators suddenly learned the utility of find -print0 and xargs -0 or find -exec rm {} +. Stéphane Chazelas for one has had a lot to say on the subject over the years.
That said, mtree (from 1989) already existed at the point that systemd-tmpfiles was invented, and already had the idea of working from a specification file with names and owners and permissions and whatnot. It is surprising that no-one ever apparently tried mtree for completely wiping /tmp. The BSDs are using find to this day in the several places with they auto-delete stuff in /tmp including the daily periodic, although they did spot that their own find had a -delete option in 1999. (-: Hell, even I am still using find.
I remember using mod_security with Apache long ago for some of this, looks like it's still around and now also supports Nginx and IIS: https://modsecurity.org/
Thank you. This doesn't have everything I'm looking for, but apparently it has been packaged in Debian at least. I don't know why the website doesn't mention this.
The Neovim plugin CodeCompanion is currently moving into a more agentic direction, it already supports an auto-submit loop with builtin tools and MCP integration.
Yes it's not a standalone CLI tool, but IMHO I'd rather have a full editor available at all times, especially one that's so hackable and lightweight.
Unfortunately, thus far, Covid19 has been through too many rapid changes for natural immunity to be effective. [0] The earlier forms allowed for it, but the evolution of the virus has outstripped most natural defences.
You got it backwards: the vaccines specifically targeted only the spike protein, while natural infection created different antibodies against all parts of the virus.
Novavax COVID-19 vaccine never targeted specifically the spike protein, and instead focused on boosting the creation of various different antibodies. It is also the least effective.
The current breed of mRNA vaccines targets the spike protein, and the TRIM21 gene. Some of them also attack the S protein directly.
All of them work to boost various antibodies. The "targets" are in addition. By targeting the spike protein, the attack hits the RNA of the virus, not just the protein. The entire cellular structure of the virus breaks down. The same with TRIM21 targets.
The bodies natural defences never targeted the spike protein, and instead focuses on the N-layer, not "all parts of the virus". These natural defences rarely manage to cause the viral cell to decay, they tend to work by slowing reproduction instead.
Novavax's vaccine was also spike-protein-only, the difference between theirs and the mRNA ones was they created it artificially in moths then extracted it for the vaccine instead of generating inside the human body.
Novavax was called "protein subunit", not "vector" - you're probably thinking of J&J and AstraZeneca which were adenovirus-vector - but also still just encoded the spike protein.
I guess fuzzing and property-based testing could mitigate this to some extent.