Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP Posting-Version: version B 2.10 5/3/83; site vaximile.UUCP Path: utzoo!watmath!clyde!cbosgd!ihnp4!houxm!vaximile!rer From: rer@vaximile.UUCP (R.RICHARDSON) Newsgroups: net.unix,net.unix-wizards,net.micro Subject: Re: Re: Binary Compatibility 80286 Message-ID: <154@vaximile.UUCP> Date: Wed, 30-Oct-85 08:16:20 EST Article-I.D.: vaximile.154 Posted: Wed Oct 30 08:16:20 1985 Date-Received: Sat, 2-Nov-85 02:10:56 EST References: <248@omen.UUCP> <10764@ucbvax.ARPA> <175@maynard.UUCP> Organization: AT&T Information Systems, Holmdel, N.J. Lines: 25 Xref: watmath net.unix:6082 net.unix-wizards:15520 net.micro:12529 > Now about this Unix source compatibility issue. Which Unix version, which > hardware, and which code are you talking about? > > I have not seen anything that demonstrates any form of compatibility between > any of the Unix' available. You have so many various forms and versions of > the ne'rdowell 'operating system' (I use the term loosely) that source written > for one can not be expected to run on any other system. > > A major cause problem of this phenomenon is to make useful software you must > take advantage of certain hardware particulars. Or you use a certain program > compiler, again making use of machine specific (or worse yet, os version > specific) routines, which again limits portability. In my experience, about 50% of the code I've run into ports to *any* UNIX on *any* box with no changes. Another 40% ports after adjusting makefiles and/or code to correct for BSD/USG specific calls, usually the tty driver stuff. And 10% ports after correcting stuff that was written unportably to begin with; it would have been unportable on *any* OS, in *any* language, on *any* box. From where I sit, the biggest win for portability would be for the BSD tty driver to disappear. Or maybe someone should write a compatibility package "ioctl" that traps and translates BSD tty ioctl's. Rick Richardson, ..!houxm!vaximile!rer