Why don't files and directories have separate namespaces?

Question

Files and directories could have different namespaces, and still be used to identify specific files, because a file and directory with the same name can be distinguished by being different kinds of things.

Primitive field and reference fields could also have different namespaces (in Java), because if a primitive and a reference field had the same name, they could be identified by being different kinds of things.

Separate namespaces are used elsewhere like this. For example, in Java, you can have a method exampleName() and a field exampleName, and though they have the same name, they are distinguished by being different kinds of things.

Answer 1

I do not believe that this would be a good idea. I imagine that the reasons involve things like performance and simplicity of the filesystem code. If a directory listing had to go down 2 or 3 or more different paths depending on how many different namespaces you think you should have, this would probably complicate the code.

Additionally, consider the end user confusion that might arise. Currently we have a kind of namespacing available in filesystems by using file extensions. You can have file.txt and file.dll and file.exe all existant in the same directory. What happens when these files are present simultaneously is a matter of concern - this has been one method for virus writers to use a form of social engineering to get you to click on the wrong file. Imagine if you could confuse a directory with a file of the same name as well?

Answer 2

First, this question is language specific. In pure OOP languages there's no distinction between atomic and compound elements. Everything is an object. By a similar reason in a pure functional language you can't have function and variable named the same.

Second, if you have polymorphic operations, there's no way to tell which variable did you refer to. For example, you can't have different namespaces for files and directories, because of the polymorphic operations, like

cp foo bar

The cp works on files and dirs, and if you have different namespaces, there's no way to tell what did you mean.

Answer 3

Directories and files aren't necessarily so different. They are both entries in their parent directory, just with a flag to indicate if an entry is a directory or not. You can open a directory and read it just as if it was a file, just that certain other operations are possible on it- symbolic links work the same way. (This description is biased towards the Unix filesystem view, but I think the DOS/Windows view works much the same way). Within any directory, there is a set of names of members, and the filesystem enforces the uniqueness constraint that a directory can only have one member with a given name.

Thinking about Java method names compared to field names--- back in C, you couldn't have a global function and global variable with the same name, because all the symbols in the object file are in a single namespace. But you could with C++, because a function "void foo()" was mapped to mangled symbol name ("foo__vv" or something). So it's not so much that they have a separate namespace, as that the key into the namespace is different for a field "foo" vs a method "foo()". Given that you can't get key clashes, they look like separate namespaces, but is that really how it's implemented?