How much is set in stone?

Steve Holden sholden at
Fri Nov 9 13:58:38 CET 2001

"Paul Rubin" <phr-n2001d at> wrote in message
news:7xofmcy0xd.fsf at
> Peter Hansen <peter at> writes:
> > A necessary step to get the robustness you claim you need is to do
> > extensive testing.  If one writes an application with Python and it
> > breaks a million dollars of merchandise it will not be Python's
> > fault, but the programmer's fault for inadequate testing.
> >
> > The only thing those other languages give you which Python doesn't
> > is compiler warnings which lull you into a false sense that your program
> > is clean when in fact you were lazy and didn't do proper testing.
> > If you do the same amount of testing with Python that you would need
> > to do with other languages to be very safe, you would have just as
> > robust an end result.
> That makes it sound as if any non-robustness in the application is the
> fault of how the application is written.  However, the Python
> interpreter itself can also have bugs.  That affects application
> robustness.  Bugs in the Python interpreter can make the application
> crash even if the application code is itself bug-free.
> Part of the question "can I write robust applications in Python" is
> "how reliable is the interpreter, really?".  I've seen Python 2.1.1
> inexplicably crash several times.  One thing that another language
> might give you besides compiler warnings and false senses of security
> is implementations that don't crash.
> Overall Python seems pretty reliable but I think its implementation is
> not yet as well shaken out as some other languages' implementations.
> I'm happy writing web applications in Python.  I could not in good
> conscience recommend writing life critical apps (medical implant
> firmware, avionics, etc.) in Python in its present form.

But of course you'd be happy to trust them to IIS-based, NT-based web
applications, right?

My own opinion, for what it's worth (and I am known to bend ears on this

    <rant>Python's delivered software quality is higher
    than many commercial language implementations, and
    the level of support available is an order of magnitude
    better: with Python you end up communicating with
    peoiple who actually understand the software</rant>.

I suspect, without any reasonable data to support it, that most such
assertions (but not necessarily yours) boil down to "I'd like to have
someone to sue if things come badly unstuck". Have you READ the typical
commercial language system license agreement?


More information about the Python-list mailing list