NetNews Usenet Archive 1992 #20/NN_1992_20.iso/spool/comp/lang/cplus/13710

Internet Message Format | 1992-09-15 | 4.9 KB

Path: sparky!uunet!sun-barr!olivea!spool.mu.edu!caen!zaphod.mps.ohio-state.edu!swrinde!elroy.jpl.nasa.gov!ucla-cs!netcon!intime.intime.COM!intime!scottm
From: [e-mail] (Scott Michel)
Newsgroups: comp.lang.c++
Subject: Re: zero-length datatype
Message-ID: <[e-mail]>
Date: 15 Sep 92 18:58:54 GMT
References: <[e-mail]> <[e-mail]>
Sender: [e-mail] (Scott Michel)
Distribution: comp
Organization: -- J/24 -- Racers Home for Programmers
Lines: 89
In-Reply-To: [e-mail]'s message of Mon, 14 Sep 92 15:26:27 GMT

In article <[e-mail]> [e-mail] (David (Lujun) Shang) writes:

   Newsgroups: comp.lang.c++
   Path: intime.intime.COM!netcon!ucla-cs!elroy.jpl.nasa.gov!sdd.hp.com!caen!uunet!panther!mothost!motsrd!news
   From: [e-mail] (David (Lujun) Shang)
   Sender: [e-mail]
   Reply-To: [e-mail]
   Organization: Motorola, Inc., Software Research and Development, Rolling Meadows, IL. 60008
   References: <[e-mail]>
   Distribution: comp
   Date: Mon, 14 Sep 92 15:26:27 GMT
   Lines: 39


Why are you arguing with what I posted? You've said the same thing I
have, but are trying to argue that we differ somehow. Granted there's
a language barrier here, but you've essentially said that same thing
I have, primarily that zero-length or no state objects don't make any
philosophical sense.

   In article <[e-mail]> [e-mail] (Scott
   Michel) writes:
   > However, it's clear that the reason why C++ doesn't allow one to have
   > a zero length object is that all instances of objects must be unique.

   An empty object has no state. You are not necessary to distinguish one
   empty object from the other. If a class can instantiate distinguished
   objects, it must be a non-empty class.

I hope I am necessary to distinguish one object from another, otherwise,
I'd also be a non-entity too. :-)

   > This is a Reasonable Thing, otherwise you'd end up with real confusion
   > in the language as to whether or not an object is really a stand-in or
   > alias for another, and whether or not the object exists in the first
   > place.

   Since you are not neccessary to distinguish non-state objects, you are
   not neccessary to distinguish whether a non-state object is a stand-in
   or alias. With zero-length assumption, you can still use object address
   in implementation to judge whether such an object exists (newed with
   initialization). But again, you can not, nor need, using object address
   to ditinguish them from one to other.

I'm not sure I understand your point here. One the one hand, you're
telling me that I need to make zero length objects be unique instances,
but similtaneously, they need not be unique because I cannot look at
the instance reference (address) to distinguish instances? I'm confused.
(Another "A can be non-A similtaneously" pitfall.)

   > Is one zero-length object itself or a meta object for another?

   C++ has not meta-object. If an object has an ditinguishable meta object,
   it does have states.

Precisely why a zero-length object makes no sense. The only place I've
found a zero length object to be useful is as a nexus of some sort or
another, which would then cause me to question whether or not I have
designed my classes correctly or not. Something without substance doesn't
have much relevance, does it? Except as a meta-something?

   > According to Aristotle's theory of classes (and his philosophy as a whole)
   > you'd force yourself into a situation where something can be itself and
   > not itself similtaneously.

   In fact, an empty class only has a single object instance. Therefore,
   when I say we need not distinguish non-state objects, they are still
   distinguishable. It does not violate Aritotle's class thoery.  We
   should distinguish the storage that holds the object and the object
   itself. We cannot distinguish non-state object through its storage,
   since it occupies no storage.

   David Shang

Where did that assumption arise, that an empty class has a single object
instance? ARE you saying is that we need not distinguish between
zero-length (or no state) objects and state objects, because they belong
to the generic class called object? It is the instances of the objects that
has everyone worried, and the semantic consequence of that fiat decision.
Secondly, what you've said (again for those of you who missed it the first
time 'round) is that "non-state objects can exist even though they don't
exist with any substance, and there's need to distinguish between one
non-state object and another even though they exist as one single instance."
That's how I read your statement, which does violate the Theory of Classes.
It got you into the pitfall that A CAN BE NON-A AT THE SAME TIME.

Now go away before I taunt you a second time...

-scottm