Typing system vs. Java

Andy Freeman anamax at earthlink.net
Sat Aug 11 00:34:42 CEST 2001


Donn Cave <donn at drizzle.com> wrote in message news:<997465342.347061 at yabetcha.sttl.drizzle.com>...
> Quoth anamax at earthlink.net (Andy Freeman):
> | Well, if we're restricting ourselves to statically typed mainstream
> | languages, they're all basically C for the purposes of this discussion.
> 
> From the "shootout" figures, OCaml would be one place to look.  This
> is not at all an endorsement of OCaml over Python, OCaml in general,
> the shootout benchmarks, an assertion that OCaml is a mainstream
> language, etc.  It does appear to be a faster option, without C's
> liabilities.

I mentioned "mainstream" because niche languages (python, ocaml, etc) don't
become mainstream by through superiority over other niche languages.  They
become mainstream through superiority over mainstream languages on important
applications.

> When we came in, if I recall the early posts to this long thread,
> someone made everyone upset by suggesting that Python's lack of
> static typing was a liability for acceptance in the industry.
> I don't know if it's true, you may or may not agree.  It isn't
> worth rehashing that all over again, but the performance benefit
> was only a beneficial potential side effect - some people actually
> have the demented view that static typing is an asset in itself.

I'd agree that static typing's benefits wrt program comprehension
exceed its benefits wrt performance, but that wasn't one of my options.

Frankly, I doubt that any of the "industry pushback" is actually
because of typing or performance, even when they're stated as a reason.
 
> | So, how many of YOUR applications would be significantly improved
> | if they were 2x faster?  Be specific?  What would you give up for
> | that speed?
> 
> Python is a good choice for the things that Python is good for.

My question has nothing to do with whether those applications are
written in python.  It is aimed at determining the relative importance
of performance.  It is aimed at determining when additional performance
would matter.

Here's another way to look at it.  Acceleration is an important aspect
of a car's performance.  Yet, people are for the most part satisfied with
cars that don't accelerate nearly as quickly as current technology can
provide, and that's been true for years.  This suggests that current
acceleration performance is "good enough" so it's time to work on other
things.

I'm suggesting that we're starting to reach the point where the same
is true of programming languages.  (The Java hit should have proved
this point, but ....)  We've reached that point wrt databases.  Yes,
there are still some applications where the relational overhead matters
(and many more where the cost of conversion is exceeds the benefits
of switching), but ....

-andy



More information about the Python-list mailing list