When MS-DOS boots up, several things have already taken place. The PC's BIOS has been powered on, a self check has been performed on the sytem's hardware, and the disk drive has already searched the master boot record and found the location of the bootable partition.
Next, a file called IO.SYS is loaded from the bootable partition. This file initializes the disk, keyboard and screen to permit input and output of data (I/O). Next, IO.SYS calls MSDOS.SYS. MS-DOS.SYS is basically the kernel of the operating system. After initializing, MS-DOS.SYS then calls COMMAND.COM. COMMAND.COM initializes then reads the batch file AUTOEXEC.BAT to determine what environment settings are required, such as the PATH variable. The PATH tells the system where to look for valid programs. If a program is not located within the path, you will have to type in the entire drive letter, path and filename information in order to launch the program.
COMMAND.COM serves as the user's 'shell' and provides the user interface to the MS-DOS 'kernel'. Anything typed at the DOS command line is sent to the COMMAND.COM interpreter.
The COMMAND.COM interpreter searches through it's memory-resident commands (commands built into COMMAND.COM itself). If it doesn't find the command, it searches through the file system in specific locations to find the transient-command within the filesystem.
Any file with the extemsions .bat, .com, .exe are considered executable, and are treated as sytem-extending 'transient-commands'. COMMAND.COM loads the command into memory, passes any additional user information to the command and begins execution.
Today, Microsoft's Windows operating systems boot in much the same way, but COMMAND.COM is bypassed and other Windows files are loaded instead.