> There's a Grade 1 bug in Microsoft's TCP/IP
> stack (PSS ID Q177074 of 12 Nov 1998) that basically means that spawning
> new
> listening sockets may fail *silently* until all sockets that were children
> of the original listening socket (all 1 of them in the current
> implementation) are closed.  No known fix.
> 
My mistake, NT Service Pack 4 fixes it.  Not that I fancy applying that to
Windows 95 :-).  My comments about writing a couple of new primitives still
apply.

		- Peter