Re: [RTnet-developers] [RFT] RT-safe TCP support for RTnet

[Vladimir Z. <vla...@si...>]

Firstly I desire to appreciate inestimable support from Jan in this
development, and I believe that even basic TCP realization in part of
RTnet will find its niche in communication solutions.

Basically the development ought to be continued, because some naturally
expected features like silly window syndrome overcoming and full blown 
server side still aren't implemented.

Hope to know any feedback of TCP in RTNet audacious usage :)

Jan Kiszka wrote:
> "TCP/IP is not supported, because it is not real-time, and there is
> no clear way to force it to be real-time.  If you don't believe me, feel
> free to try implementing TCP/IP yourself.  If necessary, an unlimited
> quantity of you-can't-do-that hubris can be provided as encouragement."
> 
> [David Schleef, developer of the original RTnet, about 10 years ago,
> http://www.rtnet.org/download/original-rtnet-0.9.1.tgz]
> 
> We are not claiming to have proven him completely wrong, but - as always
> - nothing is pure black or white. :)
> 
> 
> We are happy to announce the availability of basic but functional TCP
> support for RTnet! Check out latest git, switch on
> CONFIG_RTNET_RTIPV4_TCP, and maybe run the included examples.
> 
> Just please do not expect that everything is as "shiny" as with
> full-blown TCP stacks. The primary use case are TCP clients under RT
> constraints, connecting to existing devices that only talk this
> language. We do support the server scenario nevertheless, but accept()
> behaves a bit differently and some corner cases may not yet work perfectly.
> 
> Moreover, you have to tune your socket for sufficient packet buffers,
> specifically as you need to keep a temporary copy of each transmitted
> fragment around for the sake of retransmission on error. This is, BTW,
> limited to 3 tries so that the application is able to handle the error
> before its too late. Check out Documentation/README.tcp for more details.
> 
> Lots of credits go to Vladimir who did the hard work of developing this
> complex beast from scratch. Thanks a lot!
> 
> And I know at least one person outside Siemens is already waiting to try
> this out - so don't hesitate, we hope to hear your feedback.
> 
> Jan
> 

Vladimir