Re: a joint letter on low latency and Linux

[Phil Wilshire]

Linus et al
I've jumped in late on the debate.

Here's my 10c worth.

1..Linux is already or has the potential to become the best OS for all 
    needs. This is true for both Real Time and Non Real Time systems.

2..Why restrict its use and also restrict the applications it runs by
not    
     having a Real Time Layer ?
   It does not have to be in the standard Kernel but it should be an 
   job to add it.

3..There are Two Real Time Layers available.
   RTAI from http://www.aero.polimi.it/projects/rtai/
   RTLinux from http://www.fsmlabs.com/

   I work with both but the RTAI version has a much richer feature set.

   The features in both of them are either

       easy , very easy , to use
    or
       posix  mandated

    I disagree that either Real Time Layer is in any way shape or form
    a disease. The RTAI abstraction layer is rather neat.
    Without the RTAI module loaded the Linux Kernel is left more or less
    untouched.

4.. If you write bad programs you will write bad programs for both non
real 
    time and real time applications. If you write bad real time programs
you 
    soon find youself doing other work.

5..If you are a carefull programmer and have an absolute timing deadline
to  
   meet.   Real Time Linux is an absolute must.
   This deadline refers to a having a event happen within a few 10s of   
   milliseconds not microseconds.
   
   I am working on some of these issues and am preparing some results.
   Even the low latency  patches on an untuned disk show latencies in
the low 
   10s of  milliseconds.

Rest of thread snipped to permit more Real Time Bandwidth.

regards
    Phil Wilshire