Python for air traffic control?

Russ 18k11tm001 at sneakemail.com
Thu Jul 5 09:05:33 CEST 2001


grante at visi.com (Grant Edwards) wrote in message news:<xll07.7615$B7.1404761 at ruti.visi.com>...
> In article <bebbba07.0107022044.106d9da8 at posting.google.com>, Russ wrote:

FYI, my only access to comp.lang.python is through Google groups, and
they seem to update about once per day or so!

> >Just for the record, Mr. Park is publicly posting my private
> >replies to him. I guess I should have expected something like
> >that from him.
> 
> Ignore him.

You're absolutely right. I should have never responded to him. Note,
however, that he emailed his post directly to me. Then he edited and
posted my private replies, which were none too polite.

> >You must also understand, of course, that the safety-critical
> >application I am referring will not actually be deployed in a
> >critical mode for several years at least, during which time it
> >will be thoroughly and completely tested. Believe me, I'm
> >talking about more testing than you think. It will first be
> >used for years as a non-critical back-up for (human)
> >controllers before it is ever used as a primary means of
> >assuring separation.
> 
> My gut feeling is that if the system is done in C or C++ (for
> example) it is going to have to use dynamic memory allocation.
> Once you've taken that step, I don't think that Python is
> inherently more likely to have memory problems, and I think
> that Python is inherently _less_ likely to have many other
> types of bugs.

I have done real-time programming in C++, and I used static or
start-up memory allocation almost exclusively to avoid memory
management overhead. It really helps execution efficiency, but it
really sucks up memory. Memory is cheap these days, however. I would
probably take the same approach in Python, if possible. (As I said, I
am a Python beginner.)

> If this were a simple, hard-realtime system with response time
> requirements in the milliseconds, I'd stay away from Python,
> away from C++, and away from dynamic memory allocation.
> However, with response time requirements in the hundreds of
> milliseconds and a problem domain that probably requires
> dynamic memory allocation I think Python would be suitable.

The response time requirements are on the order of a few seconds or
so, definitely not hard real time.

> The introspection features could be very useful for automated
> testing, as could some simple instrumentation of the memory
> management functions.

Thanks for your insights.

Russ



More information about the Python-list mailing list