dn> Provided you isolate the communications side effectively in your code and
 dn> don't 'hardwire' things to the specific method being used, there's nothing
 dn> stopping you from doing this anyway.  Just create a
"communications API"
 dn> for yourself and have nothing other than code within that API use
 dn> communications specific data structures and variables; for example, use
 dn> "start_link()", "end_link()",
"check_status()", "send_data()",
 dn> "receive_data()" as generic functions, and anything which
uses those
 dn> doesn't concern itself with how the system does them. Rewriting or
 dn> replacing the communications portion becomes a lot easier as a result.

This is not as simple as it sounds.  There are a whole lot of
machine-specific stuff to be contended with.  Like can your communications
subsystem handle a receive_data() with no-wait.  What about receive_data()
with a specific number of milliseconds to wait?  What if I want to wait on
3 things - a keystroke from the user, a mouse movement from the user, or a
character from the communications device?  What if I am writing an OS/2 PM
program where the idea is that the OS tells you when something has
happened?  I haven't figured out the answers to these questions, so I
haven't got a general-purpose comms susbsystem API spec myself.  :-(  Have
you already spent the time figuring out the answers to these questions and
have something you can share with us?  :-) BFN.

Paul

--- GoldED/2 2.42.G1114