The Successor to Research Unix Was Plan 9 from Bell Labs | Hacker News Reader

The Successor to Research Unix Was Plan 9 from Bell Labs

theregister.com

31 pointsbeardyw29 comments

Plan 9 does so very much right. On Linux, namespaces are a mess of complexity which waited for Docker to make them usable; on Plan 9, a container runtime is basically a shell script.

I suspect that Plan 9 would be a more stable underlying OS for many places where Linux is used today, to include Android. I don’t know if it would be as efficient, but I really do question how much more important efficiency is than simplicity and correctness.

Linux is “correct” enough. It generally does what it says on the tin.

The sad truth is that users don’t care how simple software is to write. They do care about performance and battery life on their devices. The market chooses complex and fast over simple but slow. Every time - so long as any unnecessary complexity is hidden from users.

Simplicity buys you velocity, and velocity matters. But that’s a secondary effect.

> The market chooses complex and fast over simple but slow.

No it doesn't. Developers and their employers choose slow and complex that saves them time in development, and customers choose things that exist over things that don't. They don't normally have much of choice between simple and complex.

Do you care how simple Google search is, internally, for Google? No. All that matters is it has a good UI, and it returns good search results.

> They don't normally have much of choice between simple and complex.

I also choose IntelliJ over Vim.

As long as I don't have to buy a new computer every year to run the latest Google, I don't care. If my browser or other essential application keeps growing in ram requirements without ground breaking improvements, I grumble.

> I also choose IntelliJ over Vim.

Now do the same with your tram or parking ticket app.

Ah! But ram is a performance requirement, not a simplicity thing. Simple software can often be extremely inefficient with how it uses ram - and complex software can be complex because it makes good use of the underlying hardware.

I 100% want lean, fast software. But software being lean and fast is often at odds with it being simple. As an example, I’ve written a very complex realtime collaborative text editing library in rust that’s about 3000 loc. I’ve also ported the algorithm to JavaScript in about 500 lines. The JavaScript version is simpler by any metric - and it uses about 10x the ram and runs 100x slower than the optimised rust code.

My point is the rust code is better. It sounds like you may agree - at least in part.

Unix' popularity wasn't due to its engineering qualities, rather due to the niche it occupied in the larger ecosystem. Trying to make a slightly better Unix, but not replacing the older one in its niche wouldn't succeed. It would've been a typical "solution looking for a problem".

There are plenty of places where Linux isn't the best solution (in engineering terms), nor is anything from the Unix extended family. But there's plenty of meta-engineering factors that make it "work".

Also being freely available with source tapes helped quite a lot.

UNIX would have been hardly as successful if AT&T had been allowed to sell it from the early days.

The sad thing is that with humans momentum matters more than anything, especially logic. We are stuck with assumptions and structures from decades ago, due to the inability of most people to adapt and judge things on their merit. They subconsciously make decisions based on surface level indicators like popularity.

This is one reason that I feel that within a century or two, human-like beings will not be in control of this planet. Intelligent life will continue to evolve.

> surface level indicators like popularity

Popularity is a great heuristic for the individual - it means you'll find more examples, more tutorials, more help, more support, and others will beta test it before you get bitten. As a society, it's a prisoner's dilemma that we all keep defecting.

It is not the prisoner's dilemma. There is no third party punnishment involved.

More like "paradox of voting"?

Tragedy of the commons if you want to be pedantic, but ultimately nature is a harsh judge. Everybody loses in the end.

https://www.mdpi.com/2071-1050/13/15/8125#:~:text=A%20Rigoro.... ("Tragedy of the Commons is always a prisoner’s dilemma")

Ye sure. Maybe I could argue "nature is the third party punnisher" is a bit too reductionist, but I get your point.

> Until MS-DOS 5 in 1991, DOS didn't have a MOVE command, and no Microsoft OS had anything like symbolic links prior Windows 95's "shortcuts."

Quite the contrary, given that Microsoft Xenix was the most successful PC UNIX until Microsoft decided it wasn't worth pursuing UNIX on their portfolio.

Hardly did they knew how the server room would look like 40 years later.

Plan 9 was ported to the IBM Blue Gene supercomputer as a 'lightweight' OS.

http://doc.cat-v.org/plan_9/blue_gene/

https://www.researchgate.net/publication/225169133_Experienc...

It is extremely hard to unseat good enough and replace it with better.

Unix wasn't "good enough". It was, essentially, the standard for communication. The reason it was developed and adopted is to be used for what today we'd call a router. The idea was that big "real" computers wouldn't need to have their own networking components, instead, they'd connect to small and "dumb" computers, who, in turn would all use the same system, subsequently, would be able to connect and communicate with each other.

Unix was developed to be the small and "dumb" computer's OS. But, because it instantly monopolized the communication protocols, it became impossible to replace. Only a system that was 100% compatible with Unix would be able to do so, and that would be just another Unix.

In retrospect, it's very strange that a lot of programmers think that Unix is a good OS, or try to learn from its design. It's kind of awful, even after a lot of increments that tried to remove the awfulness and polish some rough edges, it's still really bad. And it's also getting worse, as virtually all long-living software does because of bolted-on functionality that wasn't anticipated by the original design.

Indeed, I went through a UNIX zealot phase, until the university computing archive showed me the parallel paths that failed to gain mainstream adoption.

So yeah UNIX-like shell and POSIX APIs are needed due to the amount of existing software we have out there, but that is about it, we should really move beyond it, for anything that isn't a server or embedded headless device.

Especially when “better” is expensive, but “good enough” is free.

Plan 9 is just as free as Linux.

Now, but not in the beginning, when this mattered.

Exactly - it was $500 for Plan 9 versus $0 for Linux which included all the sources until nobody was interested in Plan 9 anymore.

Just like UNIX versus VMS, MVS and others.

This is the article-format adaptation of part 3 of my FOSDEM talk earlier this month.

Part 4, the last part, coming soon, but it's more of an epilogue.