Kylix vs. gcc Development 15
Remote writes: "Borland has a paper [.pdf] by William Roetzheim comparing development and maintenance costs of software development using Kylix and gcc. Bottom line is that applications written with gcc are twice as expensive (timewise) both to write and to maintain, and about 50% more expensive to document. The comparison was done using parametric modeling techniques but the author claims that his tool has 7% accuracy. While I don't do Pascal myself, I wonder if the same would apply after they port C++ Builder to Linux and compare it to something like KDevelop + gcc."
Borland sponsored and is distributing this study -- take it for what you will. Interested parties can still produce useful studies, and may even be in the best position for collecting this sort of data. But whether Kylix matches your development environment or licensing philosophy is up to you.
No data = NOT A REAL STUDY (Score:3, Interesting)
In other words... we just released this tool. Nobody has used it enough to give us real figures, but this company, who also couldn't possibly have real figures (for the same reason), thinks that we'll save you a bundle.
AUTHOR'S Credentials (Score:1)
Is that a fact? (Score:1)
And from the paper:
Seven percent accurate? Sounds like the most accurate statement in the paper.
Even my lowbie sorceress in Diablo is more accurate with her staff than poorly-worded results like that.
On a more serious note, Borland releasing a white-paper with data pulled from a mostly-unknown delphi application should raise flags in anyone's mind. But don't for a second think that they're the first, or the last, to do it.
Flame away (Score:2, Interesting)
Let me give you an example you can't refute : C++ is easier than assembler for nearly everything. Nearly. See the pattern here ? By definition, higher level languages should be easier to code for than low level languages. C is a rather low level language, VB is a high level language, and Pascal sits somewhere in between balancing simplicity with functionality. Kylix is cloned from Delphi, and Delphi is based upon Pascal, therefore Kylix code is _usually_ easier to write than its equivalent graphical C code.
Re:Flame away (Score:1)
I'm also wondering how many of these applications that were developed, assuming this thing wasn't just dreamt up by a PR department, were done so using pre-built components. It would obviously be quicker to build some of those applications with all the base classes done for you already, especially the socket ones. You'd probably have to do a fair bit of digging to find the exact class that you wanted using just g++ and whatever comes with your Linux distro, than you would with a point-and-click interface with a "Server Applications" tab in a pop-up-window.
As for this...
By definition, higher level languages should be easier to code for than low level languages.
Actually, by definition, higher level languages are more abstracted from the way the computer thinks. This does not always make coding easier for the programmer. And I'm not talking about the inherent flaws of, say, Visual Basic before the latest service pack. I'm just talking about someone who's been spending so long in the world of C that no other language makes sense. The more sophisticated and complicated the programming you're used to doing, the less likely you're going to want to work with a language that tries to do your thinking for you.
Re:Flame away (Score:1)
Read the paper, none of the applications were built, period. Maybe they modelled them assuming pre-built applications, but the study was on the models entered into the CostXpert Tool, not on actual data. The paper was written by someone who works for the company that makes the CostXpert Tool and published by Borland. There is no reason to assume that anything in the study is valid, since the paper states that there is no empirical data possible due to the recent release of Kylix.
See my post above.
Re:Flame away - OK (Score:1)
C ~1970
C++ ~1985
Pascal was designed with simplicity as a design goal. That trade off is a lack of functionality. It is really only used as a teaching tool b/c of this.
C++ is a robust OO language.
Anyway, my rant is that you implied that C++ is a lower-level lang, when, in fact, it is considered a higher level lang.
Your point of Pascal being simplier to write is correct, however there is much (MUCH) less one can do with it.
Re:Flame away - OK (Score:1, Interesting)
Re:Flame away - OK (Score:1)
Anyway, my rant is that you implied that C++ is a lower-level lang, when, in fact, it is considered a higher level lang.
To which I must say that it may not be wise to measure the level of a language by it's syntactic ability to make higher level abstactions (which is clearly possible with C++), but rather with how deep you can dive with it, which would naturally make C++ actually as low level language as C.
Though, I must say that C++ is a fairly odd son in the field of languages. Everything goes, where you can make high level OO code you can also stick to asm inlines making it also in a reasonable way as low level as it gets.
Re:Flame away - OK (Score:1)
So is Object Pascal.
Apples and oranges (Score:3, Insightful)
Furthermore, I suspect (from my experiences with Delphi in a previous life) that gcc will be much faster for certain types of applications. Like those with no GUI, for example. Horses for courses.
Re:Apples and oranges (Score:1)
You can write console apps in Kylix (and Delphi). So the question really comes down to how well the compiler optimizes the code.
I can't imagine that
begin
writeln('Hello, world');
end.
would compile a lot different from
main() {
printf("Hello, world\n");
}
I guess it's conceivable that gcc would optimize some things better on Linux, since most of the development is tested on that platform. By the same token, I would think that bcc32 (Borland's cross-compiler) would do better on Windows platforms.