DM>     I think the above strategy would alleviate this 
 DM> problem. if there are 
 DM> chars, read them, otherwise release the time slice. If

PF>It is better to use "Wait for something" read timeout 
PF>processing;  this way,your application is blocked in the DD 
PF>-- who is the one who knows the instant something is 
PF>avaiable.

PF>There is no need to call an ioctl to see if characters are waiting.

PF>There is no need to call DosSleep

   True.  However I believe that calling DosSleep() at the right times can
significantly reduce the amount of CPU time used in a program.  EG  If a
program is expecting a 9600bps modem to be receving data full speed then it
would be better to do a DosSleep(500) and then allow the DosRead() to get
500 characters than to do multiple reads of smaller amounts of data.
    If there is any flaw in this theory or any way I can use OS/2 API calls
to get the same benefits without the effort (I have written an algorithm to
adjust the size of the DosSleep() to the rate of data transfer) then please
let me know.


    cya
___
 X MR/2 2.0 NR X Error loading KEATING.LIE. Unable to recover COUNTRY.AUS.

--- Maximus/2 2.02