MB> Without getting off on a tangent, there is no such 
 MB> thing as a standard Unix TCP/IP implementation.  You 
 MB> have a common base with a lot of similarity across 
 MB> platforms, but you also have some really oddball 
 MB> implementations such as Xenix.  Almost all reference 
 MB> books pick one or another Unix flavor and describe its 
 MB> interface.  Even Berkeley Unix has fairly radical 
 MB> changes from one version to the next.

   Well I don't consider Xenix to be UNIX at all, from what I know of it
it's too different from the regular varieties.  I agree however that there
are differences, but generally the same code will work without too many
changes.

 MB> I suppose that what most people mean when they talk about
"Unix TCP/IP" 
 MB> these days is POSIX, and that actually is not so far 
 MB> from OS/2.  However, Jim is also very right about OS/2 
 MB> being different, especially since most Unix is not 
 MB> multi-threaded, and multiple threads are a natural way 
 MB> to do a lot of the nonsense with TCP/IP that is often 
 MB> handled in Unix by asynchronous signals and such.  
 MB> While it is possible to use TCP/IP on OS/2 in the same 
 MB> general way as Unix, throwing away multi-threading 
 MB> leads to an architecture that is not optimal.

   The difference is not so great.  Just consider a UNIX program that uses
multiple processes to be equivalent to an OS/2 program with multiple
threads and things are very similar.  What you do in UNIX with multiple
processes can almost always be done with multiple threads, but with threads
you can do some other things (like dynamicaly creating sockets and sharing
them between threads) that you can't do with processes.


   cya


--- Maximus/2 2.02