Replies

• 

  Jay 🚩 :runbsd:jaypatelani@bsd.network

  Sep 7, 2026, 4:12 PM

  @tmw about copyleft and permissive. My preference or choice was due to simplicity of NetBSD's build system easy to understand , easy to tinker with. As many have already mentioned man pages are good. Innovations like UVM, nvmm, rump kernel, pkgsrc comes to mind. There's also anti AI GPLv3 OS called Ironclad but very new and probably difficult to run on bare metal for now. And gnu community seems more influenced by big tech corporations than BSD development. But time Will tell :)

  💬 0🔄 0⭐ 0
• 

  benbenjamineskola@hachyderm.io

  Sep 7, 2026, 7:33 PM

  @tmw I have a few thoughts about this.

  Like, for one thing, BSD developers have actively chosen that licence for good (to them) reasons. It’s not as if their code is being taken without their knowledge or permission and commercialised; it’s intended that their code can be used even in proprietary codebases. (One reason I’ve heard is that it’s a good thing if propriety software includes good-quality implementations of critical things.)

  And secondly, as you allude to, I don’t know how much copyleft actually protects against the real issues facing us right now. Like even before LLMs — if a codebase was GPL, on the one hand there probably isn’t much you could do to stop Amazon using it; and on the hand they absolutely had the resources to just reimplement a compatible version of something if they wanted. (I think they did so with Elasticsearch maybe?)

  (I don’t know that LLMs realistically make reimplementing something easier — they do give the appearance of doing so though, which might be all that’s necessary to cause some damage in the short term.)

  💬 1🔄 0⭐ 0
• 

  tmw 🌈tmw@ioc.exchange

  Sep 7, 2026, 7:46 PM

  @benjamineskola hmm, yeah, it's an interesting philosophy. overall, it seems like there's a lot of the same code, or types of code, being written and rewritten over and over again, which i presume is one of the things that has helped to popularize (and train) LLMs in the first place. so wanting to make as easy as possible to reuse code regardless of the context is understandable.

  💬 0🔄 0⭐ 0
• 

  Thomas :netbsd: :freebsd:tfb@bsd.cafe

  Sep 7, 2026, 8:20 PM

  @tmw The one place in the Linix kernel where I'd hope copyleft would help, would be SoC support. Anyone who's tried to run Linux on Arm quickly learns how little it helps. The vendors just take an old kernel, modify it, then make it available for download.

  Good luck trying to usefully use that, as a user. You're stuck on that precise kernel version forever.

  NetBSD's support is all pretty much upstreamed. The small vs giant project effect is important. The license changed the details of how soc vendors are shitty, but doesn't solve anything

  💬 0🔄 0⭐ 1
• 

  Fall smell is leaf mold 🍁feld@friedcheese.us

  Sep 7, 2026, 8:26 PM

  @tfb @tmw nobody really wants to contaminate their kernels with the massive amount of duplicated work to support all these stupid SoCs anyway; every single one is like supporting an entirely new arch. I've seen the discussions about it in FreeBSD and nobody wants to deal with it. NetBSD was successful because they've perfected a pretty clean way of bootstrapping up a new arch
  
  We should have had a standard like aarch64 a long time ago and this entire mess could have been avoided
  💬 0🔄 0⭐ 0