RE: CA Connect rates under RTEMS

["Jeff Hill" <johill@lanl.gov>]

> 
> BTW - the vxWorks implementation of osiSufficentSpaceInPool is of
> questionable
> usefulness so it's not really clear to me that there's any point in
> using this
> routine at all (Jeff Hill disagrees with me on this.....).
> 

My position is that it's a good idea on limited resource OS (i.e. no page
file) to try to keep at least N bytes free in pool, or a maximum free block
larger than N, and to try to prevent the CA server from using all available
memory. This is the justification for the osiSufficentSpaceInPool interface.
Whether you like it or not will depend on personal priority for robustness.

I don't disagree that so far we have not encountered a limited resource OS
that has managed to provide efficient access to such information, but that
issue IMHO is orthogonal with the justification mentioned in the previous
paragraph.

I don't doubt that a brilliant OS designer will stumble onto a solution
eventually, and until that time comes we may have to settle for a function
that always return true on certain OS, or for polling the function less
frequently (an admittedly less than satisfactory compromise). It would of
course also help to allocate *all* memory off of free lists, but this is not
easy especially with any approach to variable length strings that requires
contiguous storage.

Jeff