
Plan 9
If it ain't from outer space, I'm not interested.
Bell Labs has decided the time is right to transfer copyright of Plan 9, the not-Unix OS it created in the 1980s. Now a unit of Nokia, Bell Labs’ announcement of the shift offers a brief history of the OS. “Starting in the late 1980s, a group led by Rob Pike and UNIX co-creators Ken Thompson and Dennis Ritchie developed Plan …
A lightweight OS is to be welcomed.
Look at how the mainstream OSs have mushroomed in size.
It's almost as if the tech giants have considered Moore's Law as a licence to cram as much stuff into the OS as possible: don't worry end-user, the improved hardware that you have invested in is not for your benefit, it's for us.
Depends on just what you want from your OS.
Support for a whole raft of technology and services takes up alot of space. Universal formats to support things consumers.... Consume takes space. Admittedly support for a huge number of programming languages is unnecessary, as are including stacks to cope with unpopular tech.
Then there's the enterprise features. I mean how many home consumers need iSCSI support or the ability to support a huge raft of databases via ODBC connectors?
Then there's the issue of legacy compatibility. No-one needs to support anything older than a couple of years right?
"Support for a whole raft of technology and services takes up alot of space...including stacks to cope with unpopular tech."
The difficulty with this line of argument is that nobody wants or needs the whole set of things a modern operating system is asked to support but the subsets people do need and want can be important. Dropping them is why we find crucial and expensive pieces of technology, for instance in medical applications, dependent on out-of-date and vulnerable versions of Windows and out of production generations of computer to run it.
I agree with you in principle, but AFAIK in the specific case of medical equipment: the reason things don't get updated is that equipment is is qualified as appropriate for medical use as a whole. If you swap out a part, you need to replace it with an identical other part (or sometimes an approved replacement).
This is AFAIK. I am not an expert in this.
No. For a start it is very rare for Windows to actually drop support for anything. And, medical devices do not get their OS upgraded because then you would have to re-certify the entire machine with bodies like the FDA - a time consuming and expensive process, not something you do every 6 months.
"...Then there's the issue of legacy compatibility. No-one needs to support anything older than a couple of years right?...
I know this was tongue in cheek, but part of me responded "You'd think, right?"
Bane of my life with the types of customers I work with. Want everything modernised...good. Oh but we need to run our 20+ year old crap application the work experience kid knocked up as a bit of a project we threw their way.
"Look at how the mainstream OSs have mushroomed in size."
It depends on what you call "OS". Look at MacOS. What I would call "OS" (the Darwin kernel) is actually tiny. And then there's tons and tons and tons and more tons of software written around it. If Apple decided to replace Darwin with Plan 9, you wouldn't see any difference at all.
I remember when Plan 9 first went public.
A services-based microkernel? That was a departure from the norm, particularly the monolithic fixed-at-build-time kernels of OSes like *BSD; AIX's ability to dynamically load kernel modules was forward-thinking when it was introduced. But microkernels were already popular in the research community, and by '92 Mach was well on the way to being a true microkernel design.
The Plan 9 team made hay of the filesystem-based IPC mechanism, but we'd had essentially the same thing in UNIX for years, with UNIX-domain sockets (BSD) and the lesser-known FIFOs (AT&T V7, I think). It just wasn't so widely used. Promoting it in Plan 9 was a Good Idea, but not revolutionary.
What really made Plan 9 different was the per-process namespace. So rather than all processes seeing the same filesystem, modulo chroot(), each process would see its own version. And since pretty much everything was pushed to the filesystem as the One Abstraction to Rule Them All, that was important.
But of course in the early-to-mid-1990s it was going to be very difficult for an upstart OS to gain traction, even in the embedded market that AT&T initially targeted for commercialization. The big commercial UNIX players were fighting hard both among themselves and against various BSD distributions, against OS/2, against minicomputer platforms like the AS/400, and then against Linux and Windows. Few people wanted to bet on an outside chance. Linux and Windows were able to muscle their way into the mainstream – taking most of the market from the proprietary UNIXes – but Plan 9 really never had a chance.
There was also widespread perception that the performance cost of message-passing microkernels wasn't worth paying. This was the same period when Microsoft got rid of Windows NT's HAL, for example, and there was a general move back toward pushing all kinds of crap into kernel mode. Throw away the seatbelts to save weight.
Plan 9 might still be a nice choice for doing higher-end embedded stuff like routers, though.