C++ usage (was Re: ada and robots)

C++ usage (was Re: ada and robots)

[Brian Rogoff]

On 20 Jun 1997, Jon S Anthony wrote:
> In article <Pine.SGI.3.95.970619163029.5827A-100000@shellx.best.com> Brian Rogoff <bpr@shellx.best.com> writes:
> > > It is rather amazing, isn't it.  BTW, I've completely given up on C++.
> > > I think C still has its uses, but C++?  I don't think so.
> > 
> > I do.
> > 
> > C++ is a very widely used *family-of-languages* which will be with us for 
> > a long time.
> 
> Oh, it is _used_ all over the place (well, more or less - it's often
> difficult to glean whether it is just being used as a ANSI C sort of
> thing.)  I was talking about "uses" as in "the best thing to use here
> is C++".  This latter does not seem to have good rationale no matter
> what the situation: if it is simply C stuff - use C.  Numerical stuff
> - use Fortran.  If it is something needing flexible higher level
> capabilities, sophistication and lower level efficiencies - use Ada
> (or Eiffel).  Implemenations and availability are every bit as viable
> as those for C++.  Something even higher level - use Lisp or ST or
> some such.

My experience (which may not match yours) is that C++ is used in many 
organizations as an OO programming language, but that a subset of the 
"language" is used which is far smaller than draft ANSI/ISO C++, or 
even Lippman '91 or the "The C++ Programming Language, 2nd ed.". To be 
specific, I don't see templates, exceptions, namespaces, or RTTI being 
used much, but classes (and virtual members) and overloading are. Multiple 
inheritance is not used as much as in other languages (I've seen a
bunch of CLOS code and MI is used a lot). 
 
> > large number of projects cranking out code. So there probably is a
> > use for C++ qua C++, even if we could make technical arguments that
> > Ada can do the job "better".
> 
> I don't mean just "technically".

I don't understand what you mean here.

> > IMO, its really a question of degree only. All computer languages
> > fundamentally suck.
> 
> Right.  This I completely agree with.  However, a large enough
> quantitative gap gives a qualitative difference.

Sure. If I had my way, I'd never write another line of C++. I suppose 
Eiffel and OCAML programmers might say the same thing about Ada ;-)

> > I just happen to find Ada's flaws far more palatable for those
> > programming tasks for which C, C++, and Fortran are often used.
> 
> I hear ya.  But, for me, the degree to which C++ sucks puts it in a
> class of its own.

I have to say I mostly agree, though IMHO Perl comes close.

-- Brian

Re: C++ usage (was Re: ada and robots)

[Brian Rogoff]

On 24 Jun 1997, Jon S Anthony wrote:
> Simply that there are even simple management/economic reasons why C++
> is not a rational choice.  The really odd thing is that IME I've seen
> people actually admit this was true (that an alternative would
> actually make more sense all around), but then choose C++ anyway
> "because that is what is being used in the industry".  That's
> basically irrational.

In other words, people who had the choice admit that they made what they 
thought (before they committed) was the wrong choice? 

I'd be curious about the management/economic issues you cite. The more ammo
the better.
 
> > Sure. If I had my way, I'd never write another line of C++. I suppose 
> > Eiffel and OCAML programmers might say the same thing about Ada ;-)

> :-).  CL or ST more likely (actually I don't know OCAML - I presume it
> is functional?)  OTOH, I know a number of Lisp types who don't view
> Ada as something that sucks.  They wouldn't exactly jump at the chance
> to program in it, but they see its merits.  C++ they simply disdain...

Understandable that they might feel that way, as programming in Lisp is 
far different from programing in Ada. Lisp environments tend to really 
support an exploratory style of programming (hacking ;-) which can be 
quite useful at times. 

OCAML, which is a member of the ML family of languages, is an impure  
functional language, like Scheme, which is statically typed, has a module 
system supporting separate compilation, and an object system. Like SML, 
you don't have to explicitly type variables, as the compilers do type 
inference. For long lived code, I'm not so sure this is good, since
explicit types provide documentation for dumber programmers like me. OTOH, 
combined with an interpreter, it feels like an Ada-esque Lisp with its 
static typing. See http://pauillac.inria.fr/ocaml/ for details. 

No, I am not advocating that anyone drop Ada and rush to OCAML or SML
(but you can certainly think about chucking some of those bogus untyped 
scripting languages you use :-) 

> Well, OK you got me.  Still, in general, people wouldn't actually
> suggest using Perl for any sort of large scale programming - would
> they??

Yes. I have talked to members of several commercial projects which use 
Perl for large scale programming. One of them was switching to Java because 
the Perl code had become unmaintainable in their opinion.  

-- Brian

Re: C++ usage (was Re: ada and robots)

[Jon S Anthony]

In article <Pine.SGI.3.95.970623110733.2308B-100000@shellx.best.com> Brian Rogoff <bpr@shellx.best.com> writes:

> > > large number of projects cranking out code. So there probably is a
> > > use for C++ qua C++, even if we could make technical arguments that
> > > Ada can do the job "better".
> > 
> > I don't mean just "technically".
> 
> I don't understand what you mean here.

Simply that there are even simple management/economic reasons why C++
is not a rational choice.  The really odd thing is that IME I've seen
people actually admit this was true (that an alternative would
actually make more sense all around), but then choose C++ anyway
"because that is what is being used in the industry".  That's
basically irrational.


> Sure. If I had my way, I'd never write another line of C++. I suppose 
> Eiffel and OCAML programmers might say the same thing about Ada ;-)

:-).  CL or ST more likely (actually I don't know OCAML - I presume it
is functional?)  OTOH, I know a number of Lisp types who don't view
Ada as something that sucks.  They wouldn't exactly jump at the chance
to program in it, but they see its merits.  C++ they simply disdain...


> > > I just happen to find Ada's flaws far more palatable for those
> > > programming tasks for which C, C++, and Fortran are often used.
> > 
> > I hear ya.  But, for me, the degree to which C++ sucks puts it in a
> > class of its own.
> 
> I have to say I mostly agree, though IMHO Perl comes close.

Well, OK you got me.  Still, in general, people wouldn't actually
suggest using Perl for any sort of large scale programming - would
they??

/Jon

-- 
Jon Anthony
OMI, Belmont, MA 02178
617.484.3383
"Nightmares - Ha!  The way my life's been going lately,
 Who'd notice?"  -- Londo Mollari

Re: C++ usage (was Re: ada and robots)

[Will Rose]

Jon S Anthony (jsa@alexandria.organon.com) wrote:
[...]
: Simply that there are even simple management/economic reasons why C++
: is not a rational choice.  The really odd thing is that IME I've seen
: people actually admit this was true (that an alternative would
: actually make more sense all around), but then choose C++ anyway
: "because that is what is being used in the industry".  That's
: basically irrational.

No, it's not - it's a reasonable risk-aversive strategy.  They aren't
looking for an enormous success, but seek to avert major failure.  Since
"everybody uses C++" it must be useable, and since "everbody uses C++"
they won't be blamed for chosing it.  So "everybody uses C++".

(You can change C++ to 'Microsoft NT' in the above paragraph, and still
a correct statement.  The current popularity of Java is probably fashion,
a slightly different force, but in a year or two Java will probably also
be received wisdom.)

Will
cwr@crash.cts.com

Re: C++ usage (was Re: ada and robots)

[David Weller]

In article <867219575.393164@optional.cts.com>, Will Rose <cwr@cts.com> wrote:
>looking for an enormous success, but seek to avert major failure.  Since
>"everybody uses C++" it must be useable, and since "everbody uses C++"
>they won't be blamed for chosing it.  So "everybody uses C++".
>

Past: Nobody ever got fired for choosing IBM [Fortran|COBOL|etc.]

Present: Nobody ever gets fired for choosing Microsoft [C++|Java|etc.]

The melody's the same,
but the lyrics keep changing :-)

Re: C++ usage (was Re: ada and robots)

[Jon S Anthony]

In article <Pine.SGI.3.95.970624211546.29470A-100000@shellx.best.com> Brian Rogoff <bpr@shellx.best.com> writes:

> > is not a rational choice.  The really odd thing is that IME I've seen
> > people actually admit this was true (that an alternative would
> > actually make more sense all around), but then choose C++ anyway
> > "because that is what is being used in the industry".  That's
> > basically irrational.
> 
> In other words, people who had the choice admit that they made what they 
> thought (before they committed) was the wrong choice? 

That is correct.  But, as someone else pointed out, they would not say
that outright except among themselves.


> I'd be curious about the management/economic issues you cite. The
> more ammo the better.

Simply this: it was _clear_ that a) things would be done better and
cheaper if X were used _and_ b) that there would be happier and more
enthusiastic employees as they wanted to use X, understood why X would
make their life better, knew X (more did than not), and liked X.


> Understandable that they might feel that way, as programming in Lisp is 
> far different from programing in Ada. Lisp environments tend to really 
> support an exploratory style of programming (hacking ;-) which can be 
> quite useful at times. 

Exactly.  We use it here for much of our exploratory proof of concept
and proto-type work.


> OCAML, which is a member of the ML family of languages, is an impure  
> functional language, like Scheme, which is statically typed, has a module 
> system supporting separate compilation, and an object system. Like SML, 
> you don't have to explicitly type variables, as the compilers do type 
> inference. For long lived code, I'm not so sure this is good, since
> explicit types provide documentation for dumber programmers like me. OTOH, 
> combined with an interpreter, it feels like an Ada-esque Lisp with its 
> static typing. See http://pauillac.inria.fr/ocaml/ for details. 

Kewl.  I'll check it out.  Thanks for the tip.


> > Well, OK you got me.  Still, in general, people wouldn't actually
> > suggest using Perl for any sort of large scale programming - would
> > they??
> 
> Yes. I have talked to members of several commercial projects which use 
> Perl for large scale programming. One of them was switching to Java because 
> the Perl code had become unmaintainable in their opinion.  

And people wonder why software is in such an execrable state.
Criminey...

/Jon

-- 
Jon Anthony
OMI, Belmont, MA 02178
617.484.3383
"Nightmares - Ha!  The way my life's been going lately,
 Who'd notice?"  -- Londo Mollari