Brian Kernighan on the origins of Unix
Brian Kernighan on the origins of Unix
Posted Jan 20, 2022 9:37 UTC (Thu) by gasche (subscriber, #74946)In reply to: Brian Kernighan on the origins of Unix by karim
Parent article: Brian Kernighan on the origins of Unix
- Git made distributed version-control widely available and changed our development practices in a radical way.
- Haskell is a radical programming language that gave a lot of ideas now adopted by other languages
- Nix and Guix are proposing a fresh take on software package management and generally the OS
- QubesOS is also a fairly radical take on OS (although it is a layer on top of existing systems for usability reasons)
- the Coq proof assistant ( https://en.wikipedia.org/wiki/Coq ), and in general Interactive Theorem Provers, are radical projects and all the active ones are open source, with most run as free software projects (with a strong academic bias among contributors).
- I believe that some peer-to-peer projects also had a radical impact (Bitorrent these days), but I don't know enough about their history to tell if the open source / free software community participated actively. (I think there was a fair amount of Windows-based freeware at the time.)
- I would guess that Bitcoin was also operated by a crew of enthusiast hackers from the open source community at the start -- but again I'm not sure.
Many of those projects were born in academia (a natural place to look for radical ideas), but they were also open source projects from the start, and in many cases the free software community contributed greatly to the fact that they became successful enough to spread their ideas to the rest of the software world.
For Git one could argue that this is a commoditization of the proprietary system BitKeeper; but there were also open source ancestors (GNU Arch) and competitors (darcs, Mercurial) that were also influential.
(I think it would be worth articulating whether we are discussing inventions that were born in the open source community and became successful-enough within it, or inventions that were born anywhere but remained fairly confidential, and became successful thanks to the open source community. I rather have the later in mind, and I think it is different from "commoditization".)
One important area of innovation that is missed here is machine learning, that saw relatively little involving from well-identified open source communities, possibly as the requirements for entering the space (having a *lot* of training data at hand) made distributed development difficult. There are now striving open source projects within the machine learning software ecosystem, but I wouldn't say that the field would be substantially different without the open source community (by which I mean: without large-scale cooperation of hobbyists / benevolent contributors).
Posted Jan 20, 2022 16:55 UTC (Thu)
by karim (subscriber, #114)
[Link]
I somewhat feel git might actually a good example of commoditization. I followed this very much in "real-time" as the BitKeeper drama unfolded and was at OLS in 2005 when Matt Mackall first presented Mercurial -- I loved mercurial btw, and I was sad to see git prevail, but I digress. It was my understanding that BitKeeper was tailor-built by Larry McVoy for Linux development based on direct conversations between Linus and Larry. I don't have a reference for this, so I could be misremembering, but that was my understanding. As such, when the history section of git on Wikipedia (https://en.wikipedia.org/wiki/Git#History) states: "Support a distributed, BitKeeper-like workflow." as one of the goals set out by Linus for git, one can probably easily overlook the work that Larry presumably did to actually and actively listen to Linus and come up with something that fit "the customer's needs". [side bar: I've butted heads with Larry several times in the early 2000s on unrelated topics, but credit should go where it's deserved.]
And maybe this is what it boils down to. "Listening to the customer's needs" and "creating a product the customer will actually use" have nothing to do with the ability of replicating in an open source manor the end result of what those customer-facing steps entail. In fact, I'd venture to say that the open source community generally has had a bad history of being able to listen to customer needs. It's been, on other hand, very effective at replicating what those that have have produced, albeit sometimes in a more sustainable fashion ... because direct monetization wasn't the end goal or even possibly a need.
Again: 1) very rough ideas/arguments off the top of my head, 2) I care not of being right nor making a point. i.e. destroy/demolish at will.
Brian Kernighan on the origins of Unix