[haberman]

> The current set of namespaces in the kernel are: mount, pid, uts, ipc, net, user, and cgroup. [...] [Time is] not namespaced. [...] The kernel keyring is another item not namespaced.

I've always argued that "everything is a file" is an exaggeration. These moments make the extent of that exaggeration clear.

If everything truly was a file, the only thing you would need to namespace is the filesystem. But in reality there are a lot of other kernel objects that are not files at all.

[zapita]

You are 100% correct. “Everything is a file” was more of an early design insight, which was gradually abandoned as new features were added.

There is a movement of “Unix purists” who lament this deviation from founding principles, and advocate for a return to them. The most notable example is Plan 9.

In Plan 9, everything actually is a file. And exactly as you said, all resources are namespaced via the filesystem. It’s quite elegant and practical.

Sadly Plan 9 has remained a fringe OS, and although it influenced mainstream operating system design in many ways (including the concept of /proc), I wish that influence had been stronger.

[AceJohnny2]

I also liked QNX, when I worked with it.

You really did access devices through the /dev/ system, and device-drivers were userspace programs that created files in /dev/.

If your driver crashed, you could kill the userspace driver (which deleted the file under /dev) and restart it (assuming hardware blah blah blah).

[Someone]

”device-drivers were userspace programs that created files in /dev/. If your driver crashed, you could kill the userspace driver (which deleted the file under /dev)”

I think that shows not everything is a file. If everything were, you would start the driver by creating the file (say as a hard link from a file in /dev to the driver executable) and kill the driver by rm-ing the file.

(Chances are that, if you follow this through, this idea won’t support everything we want to do with drivers, but if so, that’s an indication that “everything is a file” doesn’t work)

[zapita]

To give you a sense of how far Plan9 took the idea... To open a tcp connection, you create a special “control file” at `/net/tcp/ctl` or some similar path, then write newline-terminated text commands to the file descriptor. That descriptor now represents your socket. You can also browse its contents as a directory (in plan9 each node in the filesystem can be both a regular file and a parent directory).

[AceJohnny2]

Great point.

[pjmlp]

It might have been elegant, but doing high performance graphics rendering wasn't something Rio was able to do.

[temac]

hm is NT a purer Unix than Unix then? After all, it has all its object in a filesystem-like tree...

[tadfisher]

In Windows NT, everything is an object. This is derived from VMS, which is essentially NT's predecessor (a principal designer of both being David Cutler of Digital Research and later Microsoft).

[jackfraser]

The problem I always had with this was that Windows has this whole layer of objects and what might even be elegance in places all hidden under the hood, and unless you're a C++ hacker you can't actually work with most of it. CMD and the GUI tools never exposed half of it to you; Powershell helps, but it's all still very hidden and hard to get to.

In comparison, Unix provides all the tools needed to take it apart and put it back together again. When you do need to interact with some syscall interface, there's almost always a complete CLI around it. It really makes it much easier to get into the nooks and crannies, inconsistencies aside.

[yjftsjthsd-h]

Yeah, as a person who hates Windows and loves all things unix, the NT kernel and underlying system have long struck me as a well-designed, nice system... with a poor userland and a terrible UI on top. But the kernel is nice.

[atombender]

He's talking about the kernel. The Windows NT kernel and the Windows APIs use handles to represent kernel/API objects, and every handle has things like security associated with it.

For example, CreateProcess() gives you back a HANDLE value representing the process, and you can close it with CloseHandle(). Everything is a HANDLE: Files, pipes, threads, etc. A notable exception is sockets, which for historical reasons use an API modeled on BSD sockets.

The object stuff you're talking about is presumably COM, which is different. COM is great, but has nothing to do with the kernel.

[hyperman1]

He's not talking about COM. Thé HANDLEs are a handleiding to a kernel object. This means multiple differentiatie handles light exist to the same object, and handles in different processen have probably differentiatie numeric values for the same object. Think win32 handle = Unix file descriptor (roughly)

[martyvis]

David Cutler developed VMS at Digital Equipment Corporation (DEC). Digital Research was a different company - it developed CP/M.

[caf]

I've always argued that "everything is a file" is an exaggeration.

This is true, but also bear in mind that "everything is a file" didn't mean "everything is represented by a name in the mount tree", it really meant "(almost) everything is referred to by a file descriptor".

It's still true that the most painful things to deal with are the ones that aren't represented by a file descriptor.

[yayana]

I've always thought of it as the preferred interface to Userland when there isn't an overriding factor.

Within a kernel it seems like no one cares how the sausage is made.

[DonHopkins]

If time were a file, you could gzip it up to compress it, and store it away for later.

Time files like an arrow!

[steffan]

Fruit files like an Apple?

[tyingq]

I agree it's unfortunate, but it doesn't really seem in conflict with "everything is a file".

Making it a file is separate from making it sensible/usable for containers. Like the /proc filesystem. They are "files", but don't many don't work as expected without something like lxcfs. Like /proc/uptime, for example.

[sytelus]

The abstraction is not really file but stream of bytes. It turns of any object with stream of bytes will need similar set of operations: open, create, read, write, close, seek etc. This is fairly generic and powerful abstraction.

[cyphar]

The abstraction is a file descriptor. Not all things represented by file descriptors support read(2) or _llseek(2), but by representing them as file descriptors you can reuse other things like af_unix file descriptor passing.

[fulafel]

Yeah, this was one of the headline changes in Plan 9 (the second OS made by the Fathers of Unix).

[pjmlp]

And improved in Inferno, which fixed some of Plan 9 flaws, the third OS made by the Fathers of Unix that HNers keep forgetting about.

[madhadron]

Everything is a file hasn't been true for Unix since almost the beginning. It's kind of like the Unix philosophy of small, independent tools...except for the database where you store all the important data.

[lunchables]

I thought "everything is a file" referred to user land.