> > really?  from susv3 (pwrite, write):
> > 
> > > This volume of IEEE Std 1003.1-2001 does not specify behavior of concurrent
> > > writes to a file from multiple processes.
> 
> hmm, I didn't know about that, that's apparently new with SUS v3.
> I'll try to dig up a specific reference to this issue in an older spec.

It might have been added in order to make NFS (or anything else that
does local RMW cycles) conform.

There might be some words about writes in O_APPEND - where you
really do want the write and the increment of the file pointer
to be atomic.

	David

-- 
David Laight: david@l8s.co.uk