I was providing an USB pen drive with the current version of the Arch Linux bootable ISO image. cp archlinux.iso /mnt/ would be pretty useless, as it would just copy the ISO file into the directory /mnt. Instead, I copied over the ISO "in raw" directly to the special device file /dev/sda, s.t. it becomes the content of the corresponding device (the pen drive here) directly, so it can be booted later on. In fact, it doesn't matter which program opens that device file /dev/sda. All the translation to the raw memory of the device is done by the kernel. So it doesn't matter if dd or cp is used. In this respect, the work is not done by dd. It is not even much faster. For both programs, /dev/sda is used as a file lying around in /dev/, which is opened in writing mode, and whose content is overwritten.

 

At this point, the ingenuity of "device files" becomes apparent. I always find it a pretty genious way of exposing devices in a filesystem tree.

Best,

Friedrich