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