Re: An Ada Advice Inquiry

[Justin Gombos <rpbkbq.xax.gld@uluv.kbq>]

On 2007-05-08, Maciej Sobczak <see.my.homepage@gmail.com> wrote:
> On 8 Maj, 04:42, Justin Gombos <rpbkbq.xax....@uluv.kbq> wrote:
>
>> An Ada implementation is considerably more complex.
>
> This statement is inconsistent with the usual criticism of C++ as
> the most difficult to parse, with too many exceptions and
> self-conflicting standard.

I would agree that C++ is more difficult to parse, but it's only
inconsistent with one element of the total effort.  The parsing
difficulty is only a marginal factor once you account for more
components of compiler production, like the runtime environment, and
the expressive power of the target language.

>> It takes one man year to produce a C++ compiler
>
> This statement is inconsistent with the usual criticism of C++ as
> the language that 9 years after the standard still doesn't have
> conforming compilers.

There's no inconsistency there, considering all the reasons why a C++
compiler might not conform to a standard beyond the effort of doing
so.  The first excuse that enters my mind is C++ compiler writers have
nowhere near the pressure that Ada compiler writers have to conform to
a standard.  The Ada market is very small, and an Ada compiler that
doesn't conform is nearly worthless.  And you said it yourself: C++ is
a self-conflicting standard.  That doesn't necessarily make the
compiler more complex, unless the authors choose to support multiple
conflicting rules and enable the user to choose between them - but
that's extra work it would be purely voluntary.

> It is also inconsistent with your expectation to get bug-free
> certified compilers. If making a C++ compiler is 10 times easier
> than the Ada one, then surely C++ compilers are more likely to
> produce high- quality code, no?

No.  Products produced using Ada achieve the most significant degree
of reliability not simply because the compiler is more correct, but
because the *language* enables developers to write correct high level
code.  Neglecting compiler induced errors, the C++ language alone
results in 3 times the defect rate of Ada code before the compiler
even touches it.  So the importance of a correct C++ compiler is
insulated by the fact that C++ is not the language of choice for
quality products.  The nature of Ada projects and the expectation of
the end products compell rigid conformance to standards to a much
higher degree.

> Apart from that, if you really think that 1m-y is enough to crank up
> the C++ compiler, why not start a company and sell one?

There are so many out there.  I would rather compete in a market with
less competition.  

-- 
PM instructions: do a C4esar Ciph3r on my address; retain punctuation.