Subject: | |
From: | |
Reply To: | |
Date: | Wed, 17 Jul 1996 20:24:46 EDT |
Content-Type: | text/plain |
Parts/Attachments: |
|
|
On Wed, 17 Jul 1996 13:13:21 -0500 Ed Pudlo said:
>As I understand it, NETIPC and TCP/IP are two different implementations
>of sockets and one can not substitute for the other. Is this incorrect?
There are actually three (well, two and a half) implementations of sockets:
* NetIPC (oldest and proprietary),
* BSD under MPE (limited to C/iX and MPE/iX 4.0 (?)
* BSD under Posix (either Posix version of C/iX or gcc and MPE/iX 5.0)
NetIPC is the most stable and robust within it's stated abilities. BSD
under MPE is a pretty good implementation of BSD but has it's restrictions
in implementation. BSD under Posix has even more limits (on most unix
platforms, a socket descriptor and a file descriptor are interchangeable
where they make sense, and you can access the socket like a file; not even
close on MPE).
The NetIPC discussion about "well-known sockets" alludes to BSD equivalents
and is the best source if you take that route. At least it is callable from
most languages. BSD is restricted to C (as far as I know). One really vague
difference is in "well-known" sockets (port number < 1024); in the original
MPE/V code you must be PM to open one. Later version didn't require PM, nor
does MPE/iX that I'm aware of. BSD does (unverified - going on comments from
Mike Belshe and others who ported early unix servers to posix when explaining
why the programs needed PM).
Jeff Kell <[log in to unmask]>
|
|
|