Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP Posting-Version: version B 2.10.2 9/18/84; site brl-tgr.ARPA Path: utzoo!watmath!clyde!burl!ulysses!allegra!mit-eddie!think!harvard!seismo!brl-tgr!gwyn From: gwyn@brl-tgr.ARPA (Doug Gwyn) Newsgroups: net.unix,net.unix-wizards,net.micro Subject: Re: Re: Binary Compatibility 80286 Message-ID: <2430@brl-tgr.ARPA> Date: Fri, 25-Oct-85 07:12:06 EDT Article-I.D.: brl-tgr.2430 Posted: Fri Oct 25 07:12:06 1985 Date-Received: Sat, 26-Oct-85 08:04:50 EDT References: <248@omen.UUCP> <10764@ucbvax.ARPA> <175@maynard.UUCP> <2380@brl-tgr.ARPA> <853@bbncc5.UUCP> Organization: Ballistic Research Lab Lines: 57 Xref: watmath net.unix:6035 net.unix-wizards:15450 net.micro:12493 It's truly amazing how people can extract more from my words than even I thought were there. I am glad to find out that I don't believe software should be tested before being sold to customers, etc. I hadn't realized I had these strange notions.. Back to the original subject: To get full executable binary interchangeability across different implementations of UNIX, you need: (1) identical instruction set architecture, including floating- point if it is used (this can be a real problem if the micro chip vendor provides a deficient FPU) (2) identical object file format, which constrains what one is able to do about such important system features as virtual memory, shared libraries, dynamic linking, checkpointing, multiple languages, software release control, etc. (3) identical system call interface, which constrains extensions, version updates, etc. (4) compatible virtual address space layouts (5) standard set of system utilities for use within applications Bill Gates's simplistic solution is for everyone to adopt Xenix, no more and no less. This is detrimental to the development of better systems! Standardization need not imply severe obstacles to progress, if done right. But just grabbing one particular implementation and requiring its use is not doing it right. As to the mass market, as I have pointed out there is no need for a single binary format even for a given chip set; there is ample precedent for variety in the marketplace, and in general it leads to better products being developed as a result of competition. As to UNIX's "success": UNIX has already been spectacularly successful, whether anyone has made a buck on it or not. There are other criteria of success than financial. Personally, I think that Thompson, Ritchie, et. al. and the organization that supported their work deserve to profit from it, but those who do little more than just resell what they developed don't get much sympathy from me when they complain that they're not getting as filthy rich as they want to be.