The kernel's code of conduct, one week later

You appear to have made your post without reading mine. I always wanted something like Git, after I came to understand the nature of distributed revision control. I played a significant part in achieving it. This is in no way consistent with your post.

Did you really just call me a genocidal monster? That would make you completely out of step with the notion of treating your colleagues with respect, and indeed, with reality. Perhaps you might wish to retract that.

> It would be interesting to have more info about the history of Git.

Well, I don't want to write a book about it right here and now. Most interested observers know that Linus got the idea of content hashing and other important ideas from Monotone:

https://lkml.org/lkml/2005/4/6/121

The main problem with Monotone was performance. Linus knew what to do about that. Monotone was the brainchild of Graydon Hoare. Incidentally, so is Rust. Quite the prolific, largely unsung inventor.

I knew Graydon from my time at Red Hat, where we both worked for Red Hat and met from time to time in the Toronto offices of the former Cygnus. Around that time, I was also casting around for a good design model for a distributed revision control system, as an alternative to Bitkeeper. I found Monotone, some months before the drama unfolded resulting in Linus's post above. Then, Monotone had no concept of directories, it only had a flat forest of objects with each content-hashed object indexed under the full path name.

One day in some dim hallway, possibly at Red Hat headquarters during a conference, I buttonholed Graydon and launched into a polemic about what I thought needed to be done to Monotone to make it truly useful. That was basically, elaborate the metadata design so that directories are objects too. Graydon initially found the idea offensive compared to what he perceived as his simpler and purer approach, but the next time I saw him, he informed me that he had in fact changed Monotone's metadata design along those lines, and Monotone's manifest was born. This is essentially what Linus found when he discovered Monotone a short time later.

Forgive me if I have erred slightly in some details, that was thirteen years ago. Graydon would probably be able to help me with this. Until now, he and I were the only two who knew this detail.

The story of how Git came to be is a whole lot more complicated than that. The real hero of the story is Andrew Tridgell, who set his brilliant mind to work on creating an open source Bitkeeper client. Though he did not intend it, a fact I can attest to from first hand knowledge, this work set in motion the chain of events that culminated in the author of Bitkeeper himself forcing Linus to do the right thing, and stop forcing the Linux kernel community to depend on a proprietary tool. Everybody knows the rest of the story, but these early events remain shrouded as a bit of a mystery.

For his efforts, Andrew (Tridge as we know him) was hounded out of the Linux kernel community in a most unseemly way. Linus should put that on his list of unfinished business that he needs to apologize for.

Linus needed to accept that embedding a bespoke tool astride the kernel development workflow was not going to end well. That point was controversial at the time, but with the clarity of historical perspective it is now settled fact. Instead of accepting that reality, Linus continued on the course of pushing the proprietary workflow further into the faces of those community members who could not accept it.

For the record, I was not the first to take exception to this ongoing encroachment. As I dimly recall, a group of us were discussing the Bitkeeper issue on the the #kernelnewbies channel, and I had taken the position that I could live with it, so long as I would be able to function as a kernel contributor without being forced to accept the Bitkeeper license. Then one of the channel members pointed out to me that Linus had just merged instructions for a "BK and Kernel Development Workflow" into the kernel source tree. It became immediately apparent that all of us who disagreed with the proprietary license were now well on the way to becoming second class citizens. That was a significant fraction of the community. I hope that you can see why I felt the need to do something.

Let me forestall the counterargument that I should really have been busy developing an alternative to Bitkeeper, as Linus suggested. I was doing exactly that, and I was not the only one. But I could not possibly have completed that work soon enough to present something usable to Linus, to stave off the looming disaster. And there were significant doubts about whether he would respond just by moving the goal posts in any case. Trying to read Linus's mind is a thankless task. In the end, Linus showed us how to do it: you need to be Linus. Then you just code up a quick prototype to show how you want it done and a team of talented coders descends on it to make it a reality.

As I said, we could have done the same in 2002 instead of 2005. It might have lacked the elegance of Monotone's content hashes but it would have done the same thing. The global distributed workflow that needed to be supported was already well understood, it already existed.

Maybe the result would have been even better than Git. Maybe it would have had cleaner commands, maybe it would have supported rename properly, who knows. That is speculation. But it is far from speculative to assert that we had the talent and manpower to do it. There was only one key ingredient missing: resolve from Linus.

If Linus had understood my message to him at the time, history would have taken a different and better course. I succeeded in getting his attention, but failed to convince him.

I hope that clarifies the historical context.