The whole utility of the World Wide Web is links, and for links to be unbroken is the whole purpose of technology.

The purpose of the Internet is to serve the content of links. Links are good for everything, and designed to allow something to be found from anywhere. In all this access and universality, there are a few no-nos.

Links should never use a directory name. The form http://domain/ is the only possible exception. You see, there is not an HTTP specification for the viewing of a "/." It is considered to be a file, or a "source" or "target" just as any other. The server is then at a loss for what to do.

Common behaviors:

• HTML "ls"-style listing.
• index.html
• 404 not found

The most common response is to create a listing, similar to the "ls" listing command formatted as HTML. But it requires an uncertain amount of computer time, and a repeated load to produce this response. And any time the software changes, your "directory" will change as well. The only way to produce real, consistent service for your website is never to link to a directory name. Put your files in files, and keep them that way.

Serious problems:

• Document ambiguity.
• Useless resource duplication.
• Unreliable and undefined behavior.

A directory is not a document. The forward slash, /, convention is used only to concatenate the separate directories, leading ultimately to the desired file. But the directory itself is not any ordinary kind of "file"; it is one computer's directory to some files, and is meaningless to another computer.

There is no such file as a "/"; a directory is not a valid link.