Why Unix Treats Everything as a File: Unpacking the File Descriptor Model

Unified Namespace in Unix/Linux

Unix and Linux expose regular files, device nodes, sockets, pipes, and many other resources through a single hierarchical namespace and a uniform set of system calls. By treating these heterogeneous objects as “files”, the kernel can manipulate them with minimal overhead and a consistent programming model.

File Descriptors

When a process opens a pathname, the kernel creates a file descriptor —an integer handle that refers to an open file description. The pathname is used only for the initial lookup; all subsequent I/O uses the descriptor with calls such as read(), write(), lseek(), and close(). The same mechanism applies to anonymous pipes and network sockets, which may have no pathname at all. Consequently, a more precise phrasing is “everything is a file descriptor”. Each process maintains its own descriptor table, mapping descriptor numbers to kernel objects.

Abstraction Limits

The “everything is a file” abstraction provides only generic operations (byte‑stream read/write, positioning, and control via ioctl()). Different resource types have distinct characteristics: block devices support random access, character devices are often stream‑oriented, and sockets implement network‑specific semantics. Because the abstraction is high‑level, device‑specific features are not directly accessible, and relying solely on generic file operations can incur inefficiencies. This trade‑off is inherent in any abstraction layer.