Conversation

Notices

1. Embed this notice
   Stefano Maffulli (ed@social.opensource.org)'s status on Tuesday, 28-Jan-2025 20:23:50 JST Stefano Maffulli

   The licenses of Llama and Deepseek contain clear violations of the Open Source Definition, these things are *not* Open Source at a more fundamental level.

   The new Open Source AI Definition doesn't even kick in.

   So much noise and disinformation.

   • Embed this notice
     Alexandre Oliva (lxo@snac.lx.oliva.nom.br)'s status on Tuesday, 28-Jan-2025 21:45:54 JST Alexandre Oliva

     in reply to
     the kernel Linux, OSS posterchild, already contained clear violations of the OSD a couple of years before OSI was founded, and I've never seen someone from OSI come forward to call it out. why would you expect to be taken seriously now?
     
     翠星石 likes this.
   • Embed this notice
     Alexandre Oliva (lxo@snac.lx.oliva.nom.br)'s status on Wednesday, 29-Jan-2025 01:38:38 JST Alexandre Oliva

     in reply to
     'fraid so. The earliest blobs in Linux were already there in 1996. And a (very very) few remain as object code disguised as sources in the kernel Linux to this day. But even when there were lots of them in disguise there, and they were growing much faster than the whole of the kernel, OSI was suspiciously silent.
     
     翠星石 likes this.
   • Embed this notice
     Stefano Maffulli (ed@social.opensource.org)'s status on Wednesday, 29-Jan-2025 01:38:39 JST Stefano Maffulli

     in reply to

     • Alexandre Oliva

     @lxo "a couple of years before OSI was founded", you wrote?

   • Embed this notice
     Alexandre Oliva (lxo@snac.lx.oliva.nom.br)'s status on Thursday, 30-Jan-2025 02:45:24 JST Alexandre Oliva

     in reply to
     no, I'm talking about incorporation of blatantly nonfree materials.
     
     the difference is subtle: the GPL's requirements apply to whole programs, not to whole distributions.
     
     the binary blobs included in Linux are largely believed to be "mere aggregation", not derivative works.
     
     but they're nonfree programs, whether because of obnoxious licenses, and/or for lack of source code, and they're distributed as part of the kernel Linux, which makes Linux nonfree.
     
     this has all long been known, but OSI has always turned a blind eye to it.
     
     now another relevant program comes about that incorporates nonfree materials in its making, and you find a need to go on the record about this fact.
     
     why the different takes? shouldn't you go (or have gone) on the record about Linux as well?
     
     that would give it at least a resemblance of coherence and adherence to principle IMHO.
     
   • Embed this notice
     Stefano Maffulli (ed@social.opensource.org)'s status on Thursday, 30-Jan-2025 02:45:26 JST Stefano Maffulli

     in reply to

     • Alexandre Oliva

     @lxo I'm confused. Are you talking about GPL violations according to the interpretation of the FSF?

   • Embed this notice
     Alexandre Oliva (lxo@snac.lx.oliva.nom.br)'s status on Thursday, 30-Jan-2025 03:14:21 JST Alexandre Oliva

     in reply to
     same reason why any other tarball or ISO or IMG file containing programs that are free and programs that are nonfree doesn't infringe on the GNU GPL, but isn't OSD-compliant as a whole.
     
     I'd be happy to have water or something else while you enjoy your beers; I'm not much of a fan of beer myself.
     
     anyhow, my qualm is not with you specifically, it's institutional. OSI has been silent on the matter, and now it comes out of the woods voicing similar concerns about a different program, but remaining silent about its posterchild. it doesn't look good.
     
     remember, it's not about copyright infringement: using materials to train LLMs doesn't require copyright permissions, and packaging even nonfree programs together is authorized by the GPL. it's about selective voicing of concerns over similar deviations from the definition your organization set out to guard.
     
   • Embed this notice
     Stefano Maffulli (ed@social.opensource.org)'s status on Thursday, 30-Jan-2025 03:14:22 JST Stefano Maffulli

     in reply to

     • Alexandre Oliva

     @lxo I still can't see why that'd be considered a violation of the OSD and not a license violation... I wasn't at the OSI at the time, I was on the FSFE board (or about to join it, I don't remember the exact dates). Worth discussing it over beers :)