Connecting Embedded Applications to the Internet

Where the Internet may have been a dedicated network for computer data
exchange in it’s infant years, today more and more small (non-PC based)
intelligent “machines” are connected to the Network of Networks. The
prediction is that by 2005, the amount of non-PC users on the Internet will far
exceed the amount of PC’s!
Today, a highly integrated microcontroller can handle the functional control of
these machines. To connect the microcontroller to the Internet, one must
implement a protocol stack on the device. To function, this protocol usually
requires a powerful processor, a complete operating system, and a large
amount of memory. In this case, the implementation of a stand-alone Internet
protocol stack could be a solution.
In order to design an embedded system, one must design both the hardware
around a microcontroller and the software. Depending on the target hardware
as well as the complexity of the problem, the developer has to decide whether
to use an operating system. According to market analysis, more than 75% of
the embedded applications use a proprietary or free operating system, about
25% do not use an operating system. Big operating systems usually have
Internet protocols already built-in, whereas deeply embedded systems do not
yet provide a TCP/IP (Transmission Control Protocol/Internet Protocol) stack.
However, a 8- or 16-bit microcontroller may not have enough resources for the
implementation of an operating system. Software implementations without
operating systems can be found in less complex applications often using small
microcontrollers with strict hardware and runtime restrictions. The choice of
writing standalone code is often made because of the need to optimize memory
usage, code size, and run-time behavior. Since the writing of this standalone
code is target specific, it is often not portable to other targets.

Links:   

http://www.freescale.com/files/microcontrollers/doc/app_note/AN2304.pdf