Re: Perl vs Python vs ....
bcwhite@verisim.com (Brian C. White) wrote on 02.08.96 in <[🔎] 320196B7.5AF715CA@verisim.com>:
> > > I'm sure C and Assembler fit "cryptic" too. Just think how much further
> > > advanced the computer industry would be if neither of those had ever
> > > been invented.
> >
> > And how much further would the industry be, if C had been typesafe (or
> > if some other, typesafe language had been used)? The expertise in
> > language design existed at the time, but C didn't have it.
>
> There were typesafe languages in the time of C: Pascal, Modula, etc.
>
> Where did they go? They didn't go anywhere because they aren't useful
> in real applications. Have you ever tried to write a dynamic skip-list
> in pascal?
Where did they go? They went and got some of the most popular
implementations, ever.
Ever heard of Turbo Pascal/Borland Pascal/Delphi? (The names change, but
it's essentially the same language.) When I have to program in the DOS
world, I'm using that almost exclusively (both at home, and in a
commercial environment). It's great for nearly everything you can do with
C, only it's a lot easier to get right. (And yes, I have tried C in the
same environment.)
There are even clones around - I know of at least four other compilers
that claim some sort of compatibility with Borland's dialect. I'd love to
have one for Linux.
The latest version can even implement a type-safe variant of printf.
Also, sometime, have a look at Modula-3. There's a free implementation
made mostly by Dec (uses gcc as backend on most architectures).
> True, but it's more likely that much of the existing code would not have
> been written at all or would not be as functional. Maintaing poorly
> structured code is hard, but not as hard as maintaining code that was
> kludged to get around limitations in the language.
I know of no language limitations in a decent, modern Pascal-style
language - such as the Borland dialects or Modula-3 - that you'd need to
kludge around, that are not also present in C.
People who say somehing like that usually think of the original Pascal
version. Pascal-style languages have come a long, long way since then.
Modern such languages include stuff like modularization, objects and
exceptions, without having the syntactic and semantic mess of C++.
> > A language's success is typically 95% who backs it, and 5% how good it
> > is. With the masses, that is.
>
> I disagree here, and MS-Dos is a great example. It's not who backs it,
> but what. Dos was backed by tonnes of software. That's why it's still
> here. Dos does the job; or did until fairly recently.
I don't see any important difference between these two paragraphs. The
main point is that technical excellence is only a very small part in
success.
> Yes! Now define "good". Good is how useful it is, not how how nicely
> it's designed.
Here's a major diagreement: nice design is an *important* part of
usefulness.
It shows in such things as training costs, and ongoing development costs
(for example, once you know the language, not only can you find errors far
faster in a Pascal-style language than in C, there are whole classes of
errors that the compiler catches that in C you'd need a debug session to
find).
> But what is the cost of that move? How many people have to be retrained?
> What are the advantages? Do the advantages outweigh the costs?
Currently, probably not, in the case of Perl in Debian.
Let's keep an open mind, however, so that if this should change, we don't
sleep through it.
MfG Kai
Reply to: