Embedded TCP/IP stack selection

A TCP/IP stack is heavily relying on the services provided by OS. Porting a stack to a different OS is a quite an involved piece of work. Thus, I would suggest using the stack which comes with your target OS.

Vladimir Vassilevsky

DSP and Mixed Signal Design Consultant

CMX-MicroNet can be used without an OS or with an OS through a simple porting layer. This porting layer is already done for the CMX-RTX and CMX-Tiny+ OSes. CMX-MicroNet supports a wide variety of processors and compilers out of the box so getting it running should not be "an involved piece of work".

Paul Bosselaers CMX Systems

In article , Vladimir Vassilevsky writes

Quite a few TCP/IP stacks are independent of any OS.

It is required to have the dynamic memory management and the timekeeping services at the least. Also, the socket interface without multithreading does not make much sense.

VLV

The code involved to write a heap, timekeeping, and scheduler is mostly high level, and is minimal compared to the code required by a TCP/IP stack. I maintain an embedded TCP/IP stack. Porting the three items you mention is far less work than writing a new Ethernet driver.

Stephen

You know everything about your own stack, thus it is not a big deal *for you* to port it to any platform. However I would prefer a solution rather then "do it yourself" package. Why bother porting somebody else's stack when there is a TCP/IP native to OS available?

VLV

I have ported lwIP and uIP for use with FreeRTOS.org on a couple of processors, and the effort was minimal. Especially in comparison to writing the MAC driver. For the path of least resistance it might be worth considering who has the best support for your chosen processor rather than your chosen OS.

Although the OP has not mentioned other requirements, such as throughput, number of connections, protocols, etc. These would naturally narrow the selection anyway.

When resources are constrained, and the generalised OS and stack are too big. Otherwise, go ahead and use an O/S with a stack.

Other reasons are to do with having full source code and so on.

Stephen

Can you let us know your experience with uC/TCP-IP? Were you able to build and test it without upfront fees?

FWIW, the 'skyeye' project has ports of lwip on uC/OS-II as a starting point for other porting efforts; I am porting it to my favorite arch over time...

Regards,

Michael

If you are not married to the M32C then take a look at:

TCP/IP, RTOS development tools and hardware all ready to use for $99.00

Paul