Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/18/84; site mips.UUCP
Path: utzoo!linus!decvax!decwrl!Glacier!mips!mash
From: mash@mips.UUCP (John Mashey)
Newsgroups: net.arch
Subject: Re: What I miss ... (really, subscript checks, uninitialized data)
Message-ID: <202@mips.UUCP>
Date: Sat, 5-Oct-85 15:30:27 EDT
Article-I.D.: mips.202
Posted: Sat Oct  5 15:30:27 1985
Date-Received: Mon, 7-Oct-85 06:51:06 EDT
References: <796@kuling.UUCP> <2580002@csd2.UUCP> <191@graffiti.UUCP> <276@watmum.UUCP>
Organization: MIPS Computer Systems, Mountain View, CA
Lines: 37

A long sequence of dicsussion got started by:
> In article <191@graffiti.UUCP> peter@graffiti.UUCP (Peter da Silva) writes:
> >> was it not the mariner probe that was lost due to a FORTRAN subscript error?
> >What should the code do when a range-check occurs? Print out an error message
> >on ticker-tape & hang? Do nothing? A better analogy, perhaps, would be...
> >
> >	"...like practicing sailing on shore with a mechanic [safety harness]
> >and leaving it on shore come the moment..."
> >
> >...you no longer have anything to attach them to.

1) Like everything else, doing subscript-checks and uninitialzied data checks
at run-time is a tradeoff: is it worth the cost or not?  Some of the arguments
that have been going on here fall into the domain-confusion problem, most
ofwhich would go away if people preferenced what they say by noting what
domain it applies to.  For example, subscript-checking might not be worthwhile
in a PC games program.  It might be elsewhere.

2) Good compiler technology can do fairly well at giving subscript-checking and
(some) uninitialized-variabel checking for nearly free; any good global
dataflow optimizer already has most ofthe data to do this.

3) It is unrealistic to think that systems must give up and die just because
they discover an internal error. I'm sure there are many counterexamples,
but there is one obvious one whose existence is familair to most people,
i.e., the telephone system's electronic switching machines.  They commonly
use error-detection/repair strategies that have multipl levels, i.e.,
if something minor goes wrong, they don't worry too much about zapping
one conversation; as damage is progressively worse, they might work all the way
up to a full reboot (anathema).  This approach is 20 years old; a recent
reference appeared in the latest ATT Technical Journal on #5ESS.  I urge people
to study old technology and uses thereof before claiming nonexistence.
-- 
-john mashey
UUCP: 	{decvax,ucbvax,ihnp4}!decwrl!mips!mash
DDD:  	415-960-1200
USPS: 	MIPS Computer Systems, 1330 Charleston Rd, Mtn View, CA 94043