Post

> we decided that the codebase is not a derivative work that would require carrying forward the GPL license and have decided to release the code under the MIT instead. oh someone wants to get sued.

I don't even think git being licensed more permissively (eg, MPL2 or something) would be a bad thing! It would be a net positive for humanity probably. Just, like. Using AI to rewrite and relicense one of the single most-used pieces of GPLed software is actually a declaration of licensing war.

And git probably has contributors with standing to sue who might have the ideological motive to do so. Also, like. How are you not a derived work? What made you come to that conclusion? Or do you just really WANT to not be?

if you, personally, were to port git to rust by hand do you think the gpl would still apply? or is the agent the problem here in your view?

I mean, if you did it after reading the git source code (which the LLMs have definitely trained on) you'd have to exercise a LOT of care to avoid being a derived work, no?

You'd probably want to have a chinese wall just so you can prove to the lawyers that you didn't directly take and mechanically translate the original code, and this is kind of the opposite of that.

(Psst, the phrase you want that doesn't have unfortunate implications is "clean room implementation")

...Right, yes. You are of course correct.

Is there a better name for that specific technique? Not super relevant to this specific thread, just figured I'd ask if you knew since IIUC you can be cleanroom without using that specific workflow and I haven't SEEN any less unfortunate names for it yet?

clean-room is always the term I’ve heard used for it, down to the plot line where they reverse engineer the PC BIOS in Halt & Catch Fire, which I believe is more or less accurate to how the first PC compatibles were created en.wikipedia.org/wiki/Clean-r...

Yeah I think you're right and I thought that clean-room was a more general term than it actually is? Ah well.

Yeah, clean-room just means "reimplemented by someone who has has the original code described for them by someone who was code-contaminated but has not seen it themselves"! We had to clean-room part of LJ when we forked

Was it overly paranoid of us, yes, but I had the sinking realization like six weeks before launch that about 80% of my useful admin tools had been ljcomint aka the not-even-source-available repo and we were trying to be on bedrock soldi legal footing

All our most active people save one were code-contaminated, too, even if we hadn't had access in ages, which is how poor Jen had to reinvent most of our admin tools with me and Mark sitting there describing them to her, heh

(Also the nix stuff, which y'all probably won't want to take but has been an absolute godsend in for me even in its present state! If it ever becomes good enough to be trustworthy and require minimal nix interaction for ongoing maintenance I'll send a PR in.)

Nix: it's better than the alternatives, but that says more about *them* than *it*.

I need to get a new laptop (mine is having a screen problem that's going to be just as expensive to fix as a new one would be) and get my dev env set up again