NetNews Usenet Archive 1992 #26/NN_1992_26.iso/spool/comp/lang/cplus/16122

Text File | 1992-11-11 | 3.6 KB | 77 lines

Newsgroups: comp.lang.c++,bnr.lang.c++
Path: sparky!uunet!portal!ntmtv!corpgate!brtph560!bnr.ca!kimmyd
From: [e-mail] (Kimmy Posey P920)
Subject: Re: Criticisms wanted
Message-ID: <[e-mail]>
Sender: [e-mail] (Usenet News)
Organization: Bell-Northern Research, BNR-RTP
References:  <[e-mail]>
Distribution: bnr
Date: Wed, 11 Nov 1992 23:17:38 GMT
Lines: 64

In article <[e-mail]>, [e-mail] (D.S.Riches) writes:
|> ... not personal ones, although I'm sure I can get them anytime :-)
|>
|> My project is at a stage where a choice between C or C++ for the
|> implementation language is required.  Most favour C++, however there
|> are some who are adamant that C++ is the wrong thing to use.  The main
|> problem is that they are not sure why!!
|>
|> Now, I've seen many pro's for the use of C++ but not really any for the
|> cons of C++.  What I'd like is a set of reasons why C++ shouldn't be
|> used.  Preferably, I want ones which are just C++ specific and not ones
|> that can also be applied to C as well.  There is no other language to
|> be considered and so I'm not interested in suggestions for going to
|> another language.
|>
|> The type of project we're working on is Real Time Communications.  Now,
|> some of the things we have to consider are; reliability, performance,
|> determinism.
|>
|> I've seen Ian Joyner's critique of C++ and the responses to that but
|> I'd like to know about other issues such as; is C++ easier to test than
|> C, is it possible to determine which object is invoked before another
|> object and so on.
|>
|> Thanks in advance for any help,
|>
|> Dave
|>
|>
|>    Dave Riches
|>    Email:  [e-mail]
|>    Smail:  BNR Europe Ltd, London Road,
|>            Harlow, Essex. CM17 9NA.  England
|>    Phone:  [phone]
|>    Fax:    [phone]

I have faced the same questions in recent months and I think I have some
insight for you.  First of all there is really is no choice here at BNR of
what language to use when implementing an object-oriented design.  No other
language having OO constructs is available (protel-2 is not here yet) with
acceptable real-time performance other than C++.  If your doin' Embedded
Systems (which is what we make money doing) then this is a paramount concern.
The good news is that the poor C++ real-time performance myth is just that.
A myth.  I can give ya a guy in Texas that pumps real-time SONAR data through
systolic arrays of DSPs and all his code is OO C++.  There is SOME real-time
degradation but think about it.  Virtual functions get called through arrays
of functions.  You do that all the time in 'C' anyway!  Second you need to
look at your design.  If your doin' the same old procedural crap, use 'C'.
If you intend to live in the present then do an OO design and use C++.  Or
implement your OO design in 'C'.  It can be done but the language doesn't
force it on ya.  Neither does C++ for that matter.  C++ inherits the same
anarchist tendencies that 'C' has.  But it does give you the ability to
create abstract data type and build inhertance relationships and stuff.
Finally, what you really need to spend time on (if you decide on OO) is
designing your object model.  This is the heart of any OO design.  I am in
the early phases of my first OO design but I have spent mucho time talking
to people who have used the technique in the real world and all of them say
that the object model is god of the design.  Keep it shallow and wide and
you should be ok.

I'm glad to see others in BNR looking into the OO paradym.  It is my opinion
that we will either embrace it or die as a company.

KimmyD