From -786786358002688162
X-Google-Language: ENGLISH,ASCII-7-bit
X-Google-Thread: f78e5,e94695845965b249
X-Google-Attributes: gidf78e5,public
X-Google-ArrivalTime: 1992-10-28 06:07:02 PST
Newsgroups: comp.std.c++
Path: sparky!uunet!charon.amdahl.com!pacbell.com!iggy.GW.Vitalink.COM!cs.widener.edu!eff!sol.ctr.columbia.edu!caen!batcomputer!munnari.oz.au!metro!extro.ucc.su.OZ.AU!maxtal
From: maxtal@extro.ucc.su.OZ.AU (John MAX Skaller)
Subject: Re: Zero-length structures and pointer comparisons
Message-ID: <1992Oct28.185947.27216@ucc.su.OZ.AU>
Sender: news@ucc.su.OZ.AU
Nntp-Posting-Host: extro.ucc.su.oz.au
Organization: MAXTAL P/L C/- University Computing Centre, Sydney
References: <1992Oct19.055511.18826@jyu.fi> <1992Oct21.153904.13965@ucc.su.OZ.AU> <1992Oct26.170404.1904@microsoft.com>
Date: Wed, 28 Oct 1992 18:59:47 GMT
Lines: 38

In article <1992Oct26.170404.1904@microsoft.com> jimad@microsoft.com (Jim Adcock) writes:

[< on pointers .. total ordering]

>This would be a big step in the wrong direction of tying C/C++ to one
>particularly simple and naive memory model, rather than allowing C/C++ to
>be reasonably ported to all of today, yesterday's and tomorrow's computer
>architectures.  IMHO.  32-bit flat architectures are simply one stop on
>the way to the future.  There are lots of interesting CPU architectures
>that aren't 0:32.  At the very least, 16:32 and 32:32 architectures are
>becoming commonplace, as is multiprocessing.  Would people force all machines
>now and in the future into the 0:32 straight-jacket?  This would be very
>short-sighted IMHO.

	I agree but I dont see what this has to do with the issue
of whether < is required to be a total order. If all pointers on
a machine are the same length, and alisaing is not allowed,
then a bitwise comparison yields a total order. That ordering
is meaningless in itself, but as Andrew Koenig pointed out it
is useful for putting pointers in data structures requiring
a type T with a total order.
>
>On the contrary, let's head in the opposite direction, and make the
>cast from ptr to int and back again strictly implementation defined, 
>thus opening C++ up to architectures with moveable objects, and opening
>the door to reasonable compacting memory management schemes.

	I'm not sure whether casting to/from int is relevant here,
except that *if* you can do it at all, you get a total order.
If you made it implementation defined (i.e. the casts were REQUIRED
to work) then you would either no longer have a total order
or you would have to have quite big 'int' values.

-- 
;----------------------------------------------------------------------
        JOHN (MAX) SKALLER,         maxtal@extro.ucc.su.oz.au
	Maxtal Pty Ltd, 6 MacKay St ASHFIELD, NSW 2131, AUSTRALIA
;--------------- SCIENTIFIC AND ENGINEERING SOFTWARE ------------------