Experiment Proposal re Languages

Experiment Proposal re Languages

[Alan Brain]

Re Language Wars.

It's my contention that not all languages are created equal. That some
have definite advantages over others, all other things being equal. For
example, and to be non-contraversial, programming in binary is likely to
be less productive than in C, in general. And a language specifically
tailored to a problem domain is likely to be better than any
general-purpose language.

So I'd like to try an experiment.

Basically, a central authority would make up a task with a reasonable
number of FPs of complexity, and write the Requirements. The task can't
be too big, as the programming would have to be done gratis - would 40
hours work be enough (implementation, that is)?

Then, via Internet, solicit a large number of suckers - um -
professional software engineers, to write a program suite which
accomplished the task. As my wife's into model railways, something
simple could be done here as regards a genuine honest-to-Ghod target
test suite.

My questions are:
What would be useful metrics? ( SLOCs, Person-Hours, etc)
What would be a reasonable upper bound to the size of the task?
Remember, this will have to be done gratis, I can't pay a cent.
What other variables should I measure? Experience, Environment, tools,
compilers?
How many victims would be required to be statistically valid, and/or
plausible?
I'd appreciate some feedback on others about this. Would it be
worthwhile? What analysis would be useful?

As I'm an Ada afficionado, I'd really like someone of the C pursuasion
to join me in designing the requirements, to eliminate unconscious bias.
Maybe even a braod spectrum of people.

I figure that I might just get a reasonable response if I could
challenge those who were barrackers for C, C++, Ada-83, Ada-95, Eiffel,
PL-I or whatever to PROVE their contentions in an experiment.

Comments, please? 

I've put this only on comp.lang.ada. Apart from comp.lang.c,
compl.lang.eiffel, what other newsgroups should this be run on? 

----------------------      <> <>    How doth the little Crocodile
| Alan & Carmel Brain|      xxxxx       Improve his shining tail?
| Canberra Australia |  xxxxxHxHxxxxxx _MMMMMMMMM_MMMMMMMMM
---------------------- o OO*O^^^^O*OO o oo     oo oo     oo  
                    By pulling Maerklin Wagons, in 1/220 Scale

Re: Experiment Proposal re Languages

[Jonas Nygren]

I believe I saw something similar, to what Alan describes below,
in the comp.lang.functional some year(s) ago. I can't remember what
languages were included in the excersise more than C and some
functional language. The C code was never completed I believe and
the functional language, perhaps Haskell, were the most efficient
in terms of man hours spent and lines of code needed.

Perhaps somebody else have a reference to this experiment?

/jonas

Alan Brain wrote:
> 
> Re Language Wars.
> 
> It's my contention that not all languages are created equal. That some
> have definite advantages over others, all other things being equal. For
> example, and to be non-contraversial, programming in binary is likely to
> be less productive than in C, in general. And a language specifically
> tailored to a problem domain is likely to be better than any
> general-purpose language.
> 
> So I'd like to try an experiment.
<snip>
> 
> ----------------------      <> <>    How doth the little Crocodile
> | Alan & Carmel Brain|      xxxxx       Improve his shining tail?
> | Canberra Australia |  xxxxxHxHxxxxxx _MMMMMMMMM_MMMMMMMMM
> ---------------------- o OO*O^^^^O*OO o oo     oo oo     oo
>                     By pulling Maerklin Wagons, in 1/220 Scale

Re: Experiment Proposal re Languages

[Larry Kilgallen]

In article <32702431.DE6@dynamite.com.au>, Alan Brain <aebrain@dynamite.com.au> writes:

> I've put this only on comp.lang.ada. Apart from comp.lang.c,
> compl.lang.eiffel, what other newsgroups should this be run on? 

Please take steps to ensure the ongoing deliberations are not
cross-posted to multiple newgroups !

Larry Kilgallen

Re: Experiment Proposal re Languages

[Franklin Chen]

Jonas Nygren <ehsjony@ehs.ericsson.se> writes:

> I believe I saw something similar, to what Alan describes below,
> in the comp.lang.functional some year(s) ago. I can't remember what
> languages were included in the excersise more than C and some
> functional language. The C code was never completed I believe and
> the functional language, perhaps Haskell, were the most efficient
> in terms of man hours spent and lines of code needed.
> 
> Perhaps somebody else have a reference to this experiment?

You are probably referring to the report "Haskell vs. Ada vs. C++
vs. Awk vs. ... An Experiment in Software Prototyping Productivity",
available at
ftp://nebula.systemsz.cs.yale.edu/pub/yale-fp/papers/NSWC/.  There is
a link to this from
http://www.cs.yale.edu/HTML/YALE/CS/HyPlans/hudak-paul.html, which
contains other links of interest concerning Haskell and functional
programming in general.

> Alan Brain wrote:
> > 
> > Re Language Wars.
> > 
> > It's my contention that not all languages are created equal. That some
> > have definite advantages over others, all other things being equal. For
> > example, and to be non-contraversial, programming in binary is likely to
> > be less productive than in C, in general. And a language specifically
> > tailored to a problem domain is likely to be better than any
> > general-purpose language.

This observation is not necessarily in conflict with the idea of
fruitfully comparing general-purpose languages.  The report mentioned
above describes how Haskell, a general-purpose language, was used to
express a domain-specific language that was well-suited to the problem
domain.

If a general-purpose language has features that facilitate the
creation of domain-specific languages (see Paul Hudak's "Building
Domain-Specific Embedded Languages",
http://www.cs.yale.edu/HTML/YALE/CS/HyPlans/hudak-dir/position-paper.ps,
for example), and therefore increase productivity, then does this not
count as an advantage of such a general-purpose language over an ad
hoc special-purpose language?

-- 
Franklin Chen                     | chen@adi.com, http://www.adi.com/~chen/
Applied Dynamics International    | http://www.adi.com/
3800 Stone School Road            | Phone: (313) 973-1300
Ann Arbor, MI 48108-2499          | FAX: (313) 668-0012

Re: Experiment Proposal re Languages

[Michiel Perdeck]

Alan,

That is an interesting proposal. I support the point but I don't think
that a small program can demonstrate the advantages Ada has over C and
C++. It is more in the organizational and structural aspects, in it's
support for sound software engineering that Ada is better that other
languages. Not for creating small applications fast. The real benefits
will pay off only in the long run because maintenance is cheaper.
In the short run, you can only convince people by showing them that a
sound structure is also more beautiful. But not everyone is able to
see that, alas.

I would be interested in commercial examples of large projects using
Ada that succeeded. Anyone haveing such examples, please write to me.

Regards, Michiel.
michiel.perdeck@cmg.nl


Alan Brain <aebrain@dynamite.com.au> wrote:

>Re Language Wars.

>It's my contention that not all languages are created equal. That some
>have definite advantages over others, all other things being equal. For
>example, and to be non-contraversial, programming in binary is likely to
>be less productive than in C, in general. And a language specifically
>tailored to a problem domain is likely to be better than any
>general-purpose language.

>So I'd like to try an experiment.

>Basically, a central authority would make up a task with a reasonable
>number of FPs of complexity, and write the Requirements. The task can't
>be too big, as the programming would have to be done gratis - would 40
>hours work be enough (implementation, that is)?

>Then, via Internet, solicit a large number of suckers - um -
>professional software engineers, to write a program suite which
>accomplished the task. As my wife's into model railways, something
>simple could be done here as regards a genuine honest-to-Ghod target
>test suite.

>My questions are:
>What would be useful metrics? ( SLOCs, Person-Hours, etc)
>What would be a reasonable upper bound to the size of the task?
>Remember, this will have to be done gratis, I can't pay a cent.
>What other variables should I measure? Experience, Environment, tools,
>compilers?
>How many victims would be required to be statistically valid, and/or
>plausible?
>I'd appreciate some feedback on others about this. Would it be
>worthwhile? What analysis would be useful?

>As I'm an Ada afficionado, I'd really like someone of the C pursuasion
>to join me in designing the requirements, to eliminate unconscious bias.
>Maybe even a braod spectrum of people.

>I figure that I might just get a reasonable response if I could
>challenge those who were barrackers for C, C++, Ada-83, Ada-95, Eiffel,
>PL-I or whatever to PROVE their contentions in an experiment.

>Comments, please? 

>I've put this only on comp.lang.ada. Apart from comp.lang.c,
>compl.lang.eiffel, what other newsgroups should this be run on? 

>----------------------      <> <>    How doth the little Crocodile
>| Alan & Carmel Brain|      xxxxx       Improve his shining tail?
>| Canberra Australia |  xxxxxHxHxxxxxx _MMMMMMMMM_MMMMMMMMM
>---------------------- o OO*O^^^^O*OO o oo     oo oo     oo  
>                    By pulling Maerklin Wagons, in 1/220 Scale

Re: Experiment Proposal re Languages

[Michiel Perdeck]

Alan,

That is an interesting proposal. I support the point but I don't think
that a small program can demonstrate the advantages Ada has over C and
C++. It is more in the organizational and structural aspects, in it's
support for sound software engineering that Ada is better that other
languages. Not for creating small applications fast. The real benefits
will pay off only in the long run because maintenance is cheaper.
In the short run, you can only convince people by showing them that a
sound structure is also more beautiful. But not everyone is able to
see that, alas.

I would be interested in commercial examples of large projects using
Ada that succeeded. Anyone haveing such examples, please write to me.

Regards, Michiel.
michiel.perdeck@cmg.nl


Alan Brain <aebrain@dynamite.com.au> wrote:

>Re Language Wars.

>It's my contention that not all languages are created equal. That some
>have definite advantages over others, all other things being equal. For
>example, and to be non-contraversial, programming in binary is likely to
>be less productive than in C, in general. And a language specifically
>tailored to a problem domain is likely to be better than any
>general-purpose language.

>So I'd like to try an experiment.

>Basically, a central authority would make up a task with a reasonable
>number of FPs of complexity, and write the Requirements. The task can't
>be too big, as the programming would have to be done gratis - would 40
>hours work be enough (implementation, that is)?

>Then, via Internet, solicit a large number of suckers - um -
>professional software engineers, to write a program suite which
>accomplished the task. As my wife's into model railways, something
>simple could be done here as regards a genuine honest-to-Ghod target
>test suite.

>My questions are:
>What would be useful metrics? ( SLOCs, Person-Hours, etc)
>What would be a reasonable upper bound to the size of the task?
>Remember, this will have to be done gratis, I can't pay a cent.
>What other variables should I measure? Experience, Environment, tools,
>compilers?
>How many victims would be required to be statistically valid, and/or
>plausible?
>I'd appreciate some feedback on others about this. Would it be
>worthwhile? What analysis would be useful?

>As I'm an Ada afficionado, I'd really like someone of the C pursuasion
>to join me in designing the requirements, to eliminate unconscious bias.
>Maybe even a braod spectrum of people.

>I figure that I might just get a reasonable response if I could
>challenge those who were barrackers for C, C++, Ada-83, Ada-95, Eiffel,
>PL-I or whatever to PROVE their contentions in an experiment.

>Comments, please? 

>I've put this only on comp.lang.ada. Apart from comp.lang.c,
>compl.lang.eiffel, what other newsgroups should this be run on? 

>----------------------      <> <>    How doth the little Crocodile
>| Alan & Carmel Brain|      xxxxx       Improve his shining tail?
>| Canberra Australia |  xxxxxHxHxxxxxx _MMMMMMMMM_MMMMMMMMM
>---------------------- o OO*O^^^^O*OO o oo     oo oo     oo  
>                    By pulling Maerklin Wagons, in 1/220 Scale

Re: Experiment Proposal re Languages

[Mark Eichin]

indeed. One conclusion I've reached in my initial study of Ada (and
extensive experience with C++) is that Ada provides better support for
multi-developer projects, whereas C++ gives a single developer more
rope, err, leverage, directly at the expense of the reader...  a small
project like this would probably show the short-turnaround prototyping
features of any language fairly well, but give no information on
scalability.

Re: Experiment Proposal re Languages

[Matthew M. Lih]

Alan Brain wrote:

> So I'd like to try an experiment.
> 
> Basically, a central authority would make up a task with a reasonable
> number of FPs of complexity, and write the Requirements. The task can't
> be too big, as the programming would have to be done gratis - would 40
> hours work be enough (implementation, that is)?
> 
> Then, via Internet, solicit a large number of suckers - um -
> professional software engineers, to write a program suite which
> accomplished the task. As my wife's into model railways, something
> simple could be done here as regards a genuine honest-to-Ghod target
> test suite.

This has been done already, and on precisely the problem you suggest.

Prof. John McCormick of State University of New York, Plattsburgh,
has students develop a software system to control a model railroad.
When he started specifying Ada as the development language instead
of C, the success rate rose significantly. Project size approximately
15000 SLOC.

You can probably find his experience on one of the Ada sites, or
if you like I saved a copy and could repost it if there's enough
demand.


Matthew M. Lih
TRW Enterprise Solutions