Typesafety versus Worse is Better - was Re: Fwd: is there any word processing software for the pdp11?

[Eric Smith]

On Wed, Dec 3, 2014 at 9:21 AM, Sean Caron <scaron at umich.edu> wrote:
> But is there necessarily a win in turning a crack C programmer into a
> novice {Smalltalk, LISP, insert your favorite language here...}
> programmers?

Of what use is a newborn baby?  Of *course* it's of no value if you stop
at turning them into a novice. But if they're actually a *good* programmer,
they'll become proficient in another language. I would go so far as to say
that if they don't become proficient in another language, given a reasonable
chance to do so, they aren't actually a good programmer, any C programming
ability notwithstanding.

> Given a short time frame, perhaps it's more effective to spend
> those hours writing code in a language with which you're already familiar,
> rather than spending them picking up a new language...

And continue writing software that has vulnerabilities. When the tools don't
provide support for writing reliable software, reliable software mostly doesn't
get written.  For every CERT advisory of a vulnerability, there are still a
huge number that have gone undetected so far.

C advocates always claim that a good programmer will write reliable code
in C, and in a vanishingly small number of cases that's true, but the
reality is that 99.9999% of C code has bugs that either:

1) could have been caught by static analysis in a language with even
    slightly better semantics (including strong typing)

2) could have been caught at runtime if the language semantics could
    reasonably support bounds checking. In this case the software would
    still have a bug, but at worst it would result in denial of service rather
    than privilege escalation and information leakage.  (Then there's the
    problem of people that use runtime checking during software testing
    to protect their valuable test data, but turn runtime checking off for
    production, but that's an argument for another day.)

There has been a huge amount of research into how to solve those
problems in C, with only partial results. It's a difficult and perhaps
intractable problem because C is a portable assembly language, with
correspondingly low level semantics The "easy" way to solve the problem
is not coming up with bandaids for C, or trying to better train C
programmers; it is to abandon C for a reasonable language.

This doesn't happen at least in part due to managers seeing a huge
glut of C programmers on the market. When all you've got is a hammer...

Note that the same arguments apply to C++.  While C++ is no longer
a proper superset of C, it still is close to being one. C++ advocates will point
out that a lot of the new stuff in C++ is safer than the old C stuff, but
unfortunately as long as the old C stuff is in there, it gets used and
causes the same old problems. Rather than C++ being a better language
than C, it's a worse language, because it has *all* of the C defects, plus
additional new ones.